Fix restore-path main↔bleQueue deadlock and courier drop amplification

A device froze permanently in a two-phone test. Debugger stacks showed an ABBA deadlock: the main actor was in bleQueue.sync (delivery-ack send → broadcastPacket → readLinkState) while bleQueue was in main.sync (captureBluetoothStatus reading backgroundTimeRemaining). The load that lined the two edges up came from a courier-drop amplification storm: drop dedup was in-memory only while the outbox driving 120s re-deposits is persisted, so every relaunch republished the same undelivered DM as a fresh 24h relay drop and every gateway relaunch re-fetched the whole backlog — ~20 copies of one DM delivered in 40ms, each triggering decrypt + delivery + ack + handshake work. Fixes, in rank order: - Edge B (P0): captureBluetoothStatus no longer main.syncs from bleQueue; backgroundTimeRemaining is sampled on main and cached behind a lock. Invariant documented: bleQueue must NEVER sync-dispatch to main. - Edge A (P0, defense in depth): sendDeliveryAck / sendReadReceipt / sendPrivateMessage / sendNoisePayload / triggerHandshake hop to messageQueue like sendMessage, so no main-actor call path reaches readLinkState's bleQueue.sync. - Drop dedup (P1): publishedDropKeys and seenDropEventIDs persist across relaunches (new BridgeDropDedupStore, entries expire with the 24h NIP-40 drop window; wiped on panic) — one drop per message ID per 24h regardless of relaunch count. - Receiver dedup (P1): openCourierEnvelope dedups on the inner private message ID before delivery, so a duplicate copy costs one decrypt and never re-delivers, re-acks, or re-triggers a handshake. - Handshake gating (P2): queued acks initiate a Noise handshake only for reachable peers; mail from absent/rotated identities no longer turns each copy into a mesh-wide handshake flood (the ack stays queued and flushes when a session eventually establishes). - Outbox (P3): re-enqueueing a queued message ID carries over its depositedCourierKeys so resends stop re-burning the same courier slots. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
2026-07-25 01:45:20 +00:00 · 2026-07-09 17:46:59 +02:00
parent 304460ee83
commit d905452df2
11 changed files with 593 additions and 38 deletions
@@ -129,6 +129,15 @@ final class BLEService: NSObject {
    // Application state tracking (thread-safe)
    #if os(iOS)
    private var isAppActive: Bool = true  // Assume active initially
+    /// Last `UIApplication.shared.backgroundTimeRemaining` sampled on the
+    /// main thread, cached so bleQueue status logs can read it without ever
+    /// dispatching to main (see `captureBluetoothStatus` for the invariant).
+    private let backgroundTimeLock = NSLock()
+    private var _cachedBackgroundTimeRemaining: TimeInterval = .greatestFiniteMagnitude
+    private var cachedBackgroundTimeRemaining: TimeInterval {
+        backgroundTimeLock.lock(); defer { backgroundTimeLock.unlock() }
+        return _cachedBackgroundTimeRemaining
+    }
    #endif
    
    // MARK: - Core BLE Objects
@@ -176,6 +185,13 @@ final class BLEService: NSObject {

    // Ingress link tracking for duplicate and last-hop suppression
    private var ingressLinks = BLEIngressLinkRegistry()
+    // Inner message IDs of recently opened courier envelopes. Redundant
+    // copies of one message ride different envelopes (each seal uses a fresh
+    // ephemeral key, and bridge drops multiply across relays/couriers), so
+    // envelope-level dedup can't catch them; dedup on the inner ID before
+    // delivery so a duplicate costs one decrypt instead of a delivery + ack
+    // + handshake each. Owned by collectionsQueue barriers.
+    private var openedCourierMessageIDs = BoundedIDSet(capacity: TransportConfig.courierOpenedMessageIDCap)
    private let logRateLimiter = BLELogRateLimiter(defaultMinimumInterval: 5)

    private var pendingPeripheralWrites = BLEOutboundWriteBuffer()
@@ -777,7 +793,11 @@ final class BLEService: NSObject {
    }
    
    func triggerHandshake(with peerID: PeerID) {
-        initiateNoiseHandshake(with: peerID)
+        // Callers are on the main actor; the handshake broadcast sync-waits
+        // on bleQueue for link state, so hop off main first.
+        messageQueue.async { [weak self] in
+            self?.initiateNoiseHandshake(with: peerID)
+        }
    }
    
    // MARK: Noise identity/session access (narrow Transport wrappers)
@@ -932,6 +952,15 @@ final class BLEService: NSObject {

    
    func sendReadReceipt(_ receipt: ReadReceipt, to peerID: PeerID) {
+        // Hop like sendMessage: callers are often on the main actor, and the
+        // send path sync-waits on bleQueue for link state — the main thread
+        // must never block on bleQueue (see captureBluetoothStatus).
+        if DispatchQueue.getSpecific(key: messageQueueKey) == nil {
+            messageQueue.async { [weak self] in
+                self?.sendReadReceipt(receipt, to: peerID)
+            }
+            return
+        }
        let payload = BLENoisePayloadFactory.readReceipt(originalMessageID: receipt.originalMessageID)

        if noiseService.hasEstablishedSession(with: peerID) {
@@ -942,11 +971,15 @@ final class BLEService: NSObject {
                SecureLogger.error("Failed to send read receipt: \(error)")
            }
        } else {
-            // Queue for after handshake and initiate if needed
+            // Queue for after handshake; initiate only while the peer is
+            // around to answer (see sendDeliveryAck — absent senders must
+            // not turn queued acks into handshake floods).
            collectionsQueue.sync(flags: .barrier) {
                pendingNoiseSessionQueues.appendTypedPayload(payload, for: peerID)
            }
-            if !noiseService.hasSession(with: peerID) { initiateNoiseHandshake(with: peerID) }
+            if !noiseService.hasSession(with: peerID), isPeerReachable(peerID) {
+                initiateNoiseHandshake(with: peerID)
+            }
            SecureLogger.debug("🕒 Queued READ receipt for \(peerID.id.prefix(8))… until handshake completes", category: .session)
        }
    }
@@ -1395,6 +1428,15 @@ final class BLEService: NSObject {
    }
    
    func sendDeliveryAck(for messageID: String, to peerID: PeerID) {
+        // Hop like sendMessage: callers are often on the main actor, and the
+        // send path sync-waits on bleQueue for link state — the main thread
+        // must never block on bleQueue (see captureBluetoothStatus).
+        if DispatchQueue.getSpecific(key: messageQueueKey) == nil {
+            messageQueue.async { [weak self] in
+                self?.sendDeliveryAck(for: messageID, to: peerID)
+            }
+            return
+        }
        let payload = BLENoisePayloadFactory.delivered(messageID: messageID)

        if noiseService.hasEstablishedSession(with: peerID) {
@@ -1404,11 +1446,18 @@ final class BLEService: NSObject {
                SecureLogger.error("Failed to send delivery ACK: \(error)")
            }
        } else {
-            // Queue for after handshake and initiate if needed
+            // Queue for after handshake; initiate only while the peer is
+            // around to answer — couriered/bridged mail routinely arrives
+            // from absent (or rotated) identities, and every duplicate copy
+            // initiating a handshake broadcast turns one undeliverable ack
+            // into a mesh-wide flood. The queued ack flushes whenever a
+            // session eventually establishes.
            collectionsQueue.sync(flags: .barrier) {
                pendingNoiseSessionQueues.appendTypedPayload(payload, for: peerID)
            }
-            if !noiseService.hasSession(with: peerID) { initiateNoiseHandshake(with: peerID) }
+            if !noiseService.hasSession(with: peerID), isPeerReachable(peerID) {
+                initiateNoiseHandshake(with: peerID)
+            }
            SecureLogger.debug("🕒 Queued DELIVERED ack for \(peerID.id.prefix(8))… until handshake completes", category: .session)
        }
    }
@@ -2720,19 +2769,37 @@ extension BLEService {
    }

    private func logBluetoothStatus(_ context: String) {
-        bleQueue.async { [weak self] in
-            guard let self = self else { return }
-            self.captureBluetoothStatus(context: context)
-        }
+        scheduleBluetoothStatusSample(after: 0, context: context)
    }

    private func scheduleBluetoothStatusSample(after delay: TimeInterval, context: String) {
-        bleQueue.asyncAfter(deadline: .now() + delay) { [weak self] in
-            guard let self = self else { return }
-            self.captureBluetoothStatus(context: context)
+        #if os(iOS)
+        // Sample the main-actor background budget first (async hop, never a
+        // sync wait), then log from bleQueue off the cache — bleQueue must
+        // never block on main (see captureBluetoothStatus).
+        DispatchQueue.main.asyncAfter(deadline: .now() + delay) { [weak self] in
+            guard let self else { return }
+            self.refreshCachedBackgroundTimeRemaining()
+            self.bleQueue.async { self.captureBluetoothStatus(context: context) }
        }
+        #else
+        bleQueue.asyncAfter(deadline: .now() + delay) { [weak self] in
+            self?.captureBluetoothStatus(context: context)
+        }
+        #endif
    }

+    #if os(iOS)
+    /// Main thread only (reads main-actor UIApplication state).
+    private func refreshCachedBackgroundTimeRemaining() {
+        dispatchPrecondition(condition: .onQueue(.main))
+        let seconds = UIApplication.shared.backgroundTimeRemaining
+        backgroundTimeLock.lock()
+        _cachedBackgroundTimeRemaining = seconds
+        backgroundTimeLock.unlock()
+    }
+    #endif
+
    private func captureBluetoothStatus(context: String) {
        assert(DispatchQueue.getSpecific(key: bleQueueKey) != nil, "captureBluetoothStatus must run on bleQueue")

@@ -2750,11 +2817,15 @@ extension BLEService {
        }

        #if os(iOS)
-        var backgroundDescriptor = ""
-        var backgroundSeconds: TimeInterval = 0
-        DispatchQueue.main.sync {
-            backgroundSeconds = UIApplication.shared.backgroundTimeRemaining
-        }
+        // INVARIANT: bleQueue must NEVER sync-dispatch to the main thread.
+        // The main actor sync-waits on bleQueue along the send paths
+        // (readLinkState), so a main.sync here completes an ABBA deadlock —
+        // field-verified as a permanent freeze when a courier-drop storm put
+        // an ack send (main → bleQueue.sync) up against a status capture
+        // (bleQueue → main.sync). backgroundTimeRemaining is main-actor
+        // state, so it is sampled on main and cached.
+        let backgroundSeconds = cachedBackgroundTimeRemaining
+        let backgroundDescriptor: String
        if backgroundSeconds == .greatestFiniteMagnitude {
            backgroundDescriptor = " bgRemaining=∞"
        } else {
@@ -3044,6 +3115,15 @@ extension BLEService {

    
    private func sendNoisePayload(_ typedPayload: Data, to peerID: PeerID) {
+        // Hop like sendMessage: the Transport-facing wrappers (verify/vouch/
+        // group payloads) call this from the main actor, and the send path
+        // sync-waits on bleQueue for link state.
+        if DispatchQueue.getSpecific(key: messageQueueKey) == nil {
+            messageQueue.async { [weak self] in
+                self?.sendNoisePayload(typedPayload, to: peerID)
+            }
+            return
+        }
        guard noiseService.hasSession(with: peerID) else {
            // No session yet - queue the payload SYNCHRONOUSLY before initiating handshake
            // to prevent race where fast handshake completion drains empty queue
@@ -3291,6 +3371,23 @@ extension BLEService {
                SecureLogger.warning("⚠️ Courier envelope carried unsupported payload type", category: .session)
                return
            }
+            let payload = Data(typedPayload.dropFirst())
+            guard let innerMessageID = PrivateMessagePacket.decode(from: payload)?.messageID else {
+                SecureLogger.warning("⚠️ Courier envelope carried undecodable private message", category: .session)
+                return
+            }
+            // Redundant copies of one message arrive as distinct envelopes
+            // (fresh seal each: mesh couriers, bridge drops across relays),
+            // so dedup here on the inner message ID — before delivery, ack,
+            // and handshake work. A duplicate costs only the decrypt above
+            // and at most one ack ever goes out per message ID.
+            let firstOpen = collectionsQueue.sync(flags: .barrier) {
+                openedCourierMessageIDs.insert(innerMessageID)
+            }
+            guard firstOpen else {
+                SecureLogger.debug("📦 Dropping duplicate courier envelope for message \(innerMessageID.prefix(8))…", category: .session)
+                return
+            }
            // Couriered mail arrives while the sender is absent, so the UI's
            // block check can't resolve their fingerprint from a live session.
            // Gate here, where the full static key is in hand.
@@ -3306,7 +3403,6 @@ extension BLEService {
            let shortID = PeerID(publicKey: senderStaticKey)
            let isKnownOnMesh = collectionsQueue.sync { peerRegistry.info(for: shortID) != nil }
            let senderPeerID = isKnownOnMesh ? shortID : PeerID(hexData: senderStaticKey)
-            let payload = Data(typedPayload.dropFirst())
            SecureLogger.debug("📦 Opened courier envelope from \(senderPeerID.id.prefix(8))…", category: .session)
            sfMetrics?.record(.courierOpened)
            notifyUI { [weak self] in
@@ -3745,6 +3841,7 @@ extension BLEService {
    #if os(iOS)
    @objc private func appDidBecomeActive() {
        isAppActive = true
+        refreshCachedBackgroundTimeRemaining()
        // Restart scanning with allow duplicates when app becomes active
        if centralManager?.state == .poweredOn {
            centralManager?.stopScan()
@@ -3758,6 +3855,7 @@ extension BLEService {

    @objc private func appDidEnterBackground() {
        isAppActive = false
+        refreshCachedBackgroundTimeRemaining()
        // Restart scanning without allow duplicates in background
        if centralManager?.state == .poweredOn {
            centralManager?.stopScan()
@@ -3851,6 +3949,15 @@ extension BLEService {
    // MARK: Private Message Handling
    
    private func sendPrivateMessage(_ content: String, to recipientID: PeerID, messageID: String) {
+        // Hop like sendMessage: the Transport-facing wrappers call this from
+        // the main actor (router sends, favorite notifications), and the send
+        // path sync-waits on bleQueue for link state.
+        if DispatchQueue.getSpecific(key: messageQueueKey) == nil {
+            messageQueue.async { [weak self] in
+                self?.sendPrivateMessage(content, to: recipientID, messageID: messageID)
+            }
+            return
+        }
        // Sessions and wire recipient IDs are keyed by the short 16-hex form;
        // callers may pass the full 64-hex noise key (mirrors sendFilePrivate).
        let recipientID = recipientID.toShort()
@@ -81,21 +81,55 @@ final class BridgeCourierService: ObservableObject {
    private(set) var myTagsHex: Set<String> = []
    private(set) var watchedPeerTags: [(peerID: PeerID, tagsHex: Set<String>)] = []
    private(set) var pendingDrops: [(envelope: CourierEnvelope, dedupKey: String?)] = []
-    /// Message IDs already published as drops (sender-side dedup).
-    private var publishedDropKeys: BoundedIDSet
-    /// Drop event IDs already handled (multi-relay dedup).
-    private var seenDropEventIDs: BoundedIDSet
+    /// Message IDs already published as drops (sender-side dedup) and drop
+    /// event IDs already handled (multi-relay dedup). Both persist across
+    /// relaunches: relays hold drops for the full 24h NIP-40 window and the
+    /// persisted outbox keeps re-depositing, so in-memory-only dedup meant
+    /// every relaunch republished the same message as a fresh drop and every
+    /// gateway relaunch re-delivered the whole backlog (field-verified
+    /// amplification storm). Entries age out with the 24h drop window.
+    private var publishedDropKeys: ExpiringIDSet
+    private var seenDropEventIDs: ExpiringIDSet
    private var subscriptionOpen = false
    private var lastSubscribedTags: Set<String> = []
    private var refreshTimerArmed = false
    private var lastAnnounceRefresh = Date.distantPast

    private let now: () -> Date
+    private let dedupStore: BridgeDropDedupStore

-    init(now: @escaping () -> Date = Date.init) {
+    init(now: @escaping () -> Date = Date.init, dedupStore: BridgeDropDedupStore? = nil) {
        self.now = now
-        self.publishedDropKeys = BoundedIDSet(capacity: Limits.maxTrackedIDs)
-        self.seenDropEventIDs = BoundedIDSet(capacity: Limits.maxTrackedIDs)
+        self.dedupStore = dedupStore ?? BridgeDropDedupStore(persistsToDisk: !TestEnvironment.isRunningTests)
+        let snapshot = self.dedupStore.load()
+        let date = now()
+        self.publishedDropKeys = ExpiringIDSet(
+            capacity: Limits.maxTrackedIDs,
+            lifetime: CourierEnvelope.maxLifetimeSeconds,
+            entries: snapshot.publishedDropKeys,
+            now: date
+        )
+        self.seenDropEventIDs = ExpiringIDSet(
+            capacity: Limits.maxTrackedIDs,
+            lifetime: CourierEnvelope.maxLifetimeSeconds,
+            entries: snapshot.seenDropEventIDs,
+            now: date
+        )
+    }
+
+    private func persistDedup() {
+        dedupStore.save(BridgeDropDedupStore.Snapshot(
+            publishedDropKeys: publishedDropKeys.entries,
+            seenDropEventIDs: seenDropEventIDs.entries
+        ))
+    }
+
+    /// Panic wipe: forget queued drops and the persisted dedup record.
+    func wipe() {
+        pendingDrops.removeAll()
+        publishedDropKeys = ExpiringIDSet(capacity: Limits.maxTrackedIDs, lifetime: CourierEnvelope.maxLifetimeSeconds)
+        seenDropEventIDs = ExpiringIDSet(capacity: Limits.maxTrackedIDs, lifetime: CourierEnvelope.maxLifetimeSeconds)
+        dedupStore.wipe()
    }

    // MARK: - Sender role
@@ -108,14 +142,15 @@ final class BridgeCourierService: ObservableObject {
    @discardableResult
    func depositDrop(content: String, messageID: String, recipientNoiseKey: Data) -> Bool {
        guard bridgeEnabled?() ?? false else { return false }
-        guard !publishedDropKeys.contains(messageID) else { return false }
+        guard !publishedDropKeys.contains(messageID, now: now()) else { return false }
        guard let envelope = sealEnvelope?(content, messageID, recipientNoiseKey) else { return false }
        // An envelope that can't encode within the drop size caps fails the
        // same way on every attempt (size is a function of the content, not
        // of the sealing); consume the dedup slot so the retry sweep stops
        // re-running Noise sealing on a drop that can never ship.
        guard let encoded = envelope.encode(), encoded.count <= Limits.maxDropEnvelopeBytes else {
-            publishedDropKeys.insert(messageID)
+            publishedDropKeys.insert(messageID, now: now())
+            persistDedup()
            return false
        }
        // Only consume the sender-side dedup slot once the drop is durably
@@ -124,7 +159,8 @@ final class BridgeCourierService: ObservableObject {
        // router's retry sweep can attempt a fresh deposit rather than
        // marking the message "carried" and blocking retries forever.
        guard publishDrop(envelope, messageID: messageID) else { return false }
-        publishedDropKeys.insert(messageID)
+        publishedDropKeys.insert(messageID, now: now())
+        persistDedup()
        return true
    }

@@ -153,7 +189,10 @@ final class BridgeCourierService: ObservableObject {
                let evicted = pendingDrops.removeFirst()
                // The oldest queued drop is being dropped before it ever
                // published; release its dedup slot so it stays retryable.
-                if let key = evicted.dedupKey { publishedDropKeys.remove(key) }
+                if let key = evicted.dedupKey {
+                    publishedDropKeys.remove(key)
+                    persistDedup()
+                }
            }
            return true
        }
@@ -180,7 +219,10 @@ final class BridgeCourierService: ObservableObject {
        for item in queued where !publishDrop(item.envelope, messageID: item.dedupKey) {
            // Compose failed with relays up: release the slot so the router's
            // retry sweep can attempt a fresh deposit.
-            if let key = item.dedupKey { publishedDropKeys.remove(key) }
+            if let key = item.dedupKey {
+                publishedDropKeys.remove(key)
+                persistDedup()
+            }
        }
    }

@@ -265,7 +307,8 @@ final class BridgeCourierService: ObservableObject {
    func handleDropEvent(_ event: NostrEvent) {
        guard bridgeEnabled?() ?? false else { return }
        guard event.kind == NostrProtocol.EventKind.courierDrop.rawValue else { return }
-        guard seenDropEventIDs.insert(event.id) else { return }
+        guard seenDropEventIDs.insert(event.id, now: now()) else { return }
+        persistDedup()
        guard let data = Data(base64Encoded: event.content),
              data.count <= Limits.maxDropEnvelopeBytes,
              let envelope = CourierEnvelope.decode(data),
@@ -0,0 +1,133 @@
+//
+// BridgeDropDedupStore.swift
+// bitchat
+//
+// This is free and unencumbered software released into the public domain.
+// For more information, see <https://unlicense.org>
+//
+
+import BitLogger
+import Foundation
+
+/// ID set with per-entry timestamps: entries expire after `lifetime` and the
+/// oldest are evicted past `capacity`. The bridge's dedup caches use this
+/// instead of `BoundedIDSet` because their contents persist across relaunches
+/// and must age out with the 24h drop window they guard.
+struct ExpiringIDSet {
+    private(set) var entries: [String: Date]
+    let capacity: Int
+    let lifetime: TimeInterval
+
+    init(capacity: Int, lifetime: TimeInterval, entries: [String: Date] = [:], now: Date = Date()) {
+        self.capacity = capacity
+        self.lifetime = lifetime
+        self.entries = entries
+        prune(now: now)
+    }
+
+    func contains(_ id: String, now: Date) -> Bool {
+        guard let recorded = entries[id] else { return false }
+        return now.timeIntervalSince(recorded) <= lifetime
+    }
+
+    /// Returns false when the ID was already present (and unexpired).
+    @discardableResult
+    mutating func insert(_ id: String, now: Date) -> Bool {
+        guard !contains(id, now: now) else { return false }
+        entries[id] = now
+        prune(now: now)
+        return true
+    }
+
+    /// Releases a previously inserted ID so it can be re-added later (e.g. a
+    /// queued drop evicted before it ever published must become retryable).
+    mutating func remove(_ id: String) {
+        entries.removeValue(forKey: id)
+    }
+
+    private mutating func prune(now: Date) {
+        if entries.contains(where: { now.timeIntervalSince($0.value) > lifetime }) {
+            entries = entries.filter { now.timeIntervalSince($0.value) <= lifetime }
+        }
+        let overflow = entries.count - capacity
+        guard overflow > 0 else { return }
+        for (id, _) in entries.sorted(by: { $0.value < $1.value }).prefix(overflow) {
+            entries.removeValue(forKey: id)
+        }
+    }
+}
+
+/// Disk persistence for the bridge courier's drop-dedup record. Relays hold
+/// drops for the full 24h NIP-40 window and redeliver them on every launch,
+/// and the 120s outbox sweep re-deposits anything undelivered — so with
+/// in-memory-only dedup every relaunch republished the same message as a
+/// fresh drop (fresh throwaway seal, undeduplicatable downstream) and every
+/// gateway relaunch re-delivered the whole backlog. Field-verified: ~20
+/// copies of one DM delivered in 40ms fed the storm behind a permanent
+/// device freeze. Persisting both sides caps this at one drop per message
+/// ID per 24h regardless of relaunch count.
+///
+/// Contents are opaque IDs (message UUIDs, relay event IDs) — no plaintext,
+/// no peer identities — so until-first-unlock protection matches
+/// `NostrProcessedEventStore`, and the file must load during a
+/// locked-background restoration relaunch. Wiped on panic with the rest of
+/// the courier state.
+final class BridgeDropDedupStore {
+    struct Snapshot: Codable {
+        var publishedDropKeys: [String: Date]
+        var seenDropEventIDs: [String: Date]
+    }
+
+    private let fileURL: URL?
+
+    /// - Parameter fileURL: Overrides the on-disk location (tests). Ignored
+    ///   when `persistsToDisk` is false.
+    init(persistsToDisk: Bool = true, fileURL: URL? = nil) {
+        self.fileURL = persistsToDisk ? (fileURL ?? Self.defaultFileURL()) : nil
+    }
+
+    func load() -> Snapshot {
+        guard let fileURL,
+              let data = try? Data(contentsOf: fileURL),
+              let snapshot = try? JSONDecoder().decode(Snapshot.self, from: data) else {
+            return Snapshot(publishedDropKeys: [:], seenDropEventIDs: [:])
+        }
+        return snapshot
+    }
+
+    func save(_ snapshot: Snapshot) {
+        guard let fileURL else { return }
+        do {
+            try FileManager.default.createDirectory(
+                at: fileURL.deletingLastPathComponent(),
+                withIntermediateDirectories: true
+            )
+            let data = try JSONEncoder().encode(snapshot)
+            var options: Data.WritingOptions = [.atomic]
+            #if os(iOS)
+            options.insert(.completeFileProtectionUntilFirstUserAuthentication)
+            #endif
+            try data.write(to: fileURL, options: options)
+        } catch {
+            SecureLogger.error("Failed to persist bridge drop dedup record: \(error)", category: .session)
+        }
+    }
+
+    /// Panic wipe: forget which drops we published or handled.
+    func wipe() {
+        guard let fileURL else { return }
+        try? FileManager.default.removeItem(at: fileURL)
+    }
+
+    private static func defaultFileURL() -> URL? {
+        guard let base = try? FileManager.default.url(
+            for: .applicationSupportDirectory,
+            in: .userDomainMask,
+            appropriateFor: nil,
+            create: true
+        ) else { return nil }
+        return base
+            .appendingPathComponent("courier", isDirectory: true)
+            .appendingPathComponent("bridge-drop-dedup.json")
+    }
+}
@@ -351,9 +351,15 @@ final class MessageRouter {
    }

    private func enqueue(_ message: QueuedMessage, for peerID: PeerID) {
+        var message = message
        var queue = outbox[peerID] ?? []
-        // Re-sending an already-queued ID replaces the entry (keeps attempt count fresh)
-        queue.removeAll { $0.messageID == message.messageID }
+        // Re-sending an already-queued ID replaces the entry (keeps attempt
+        // count fresh) but must not forget which couriers already carry it,
+        // or the replacement re-burns the same courier slots.
+        if let existing = queue.firstIndex(where: { $0.messageID == message.messageID }) {
+            message.depositedCourierKeys.formUnion(queue[existing].depositedCourierKeys)
+            queue.remove(at: existing)
+        }
        queue.append(message)

        // Enforce per-peer size limit with FIFO eviction
@@ -345,6 +345,9 @@ enum TransportConfig {
    // Cooldown between speculative multi-hop handovers of the same envelope
    // toward a recipient heard only via relayed announces.
    static let courierRemoteHandoverCooldownSeconds: TimeInterval = 10 * 60
+    // Recently opened courier inner message IDs kept for receiver-side dedup
+    // (redundant copies ride distinct seals, so only the inner ID matches).
+    static let courierOpenedMessageIDCap: Int = 512

    // One-time prekey bundles (forward-secret courier sealing)
    // Own gossip-sync round for bundles: modest cadence, bounded peer count,
@@ -1173,6 +1173,7 @@ final class ChatViewModel: ObservableObject, BitchatDelegate, TransportEventDele
        // our own queued outbox, the carried public history, and the
        // counters describing all of it
        CourierStore.shared.wipe()
+        BridgeCourierService.shared.wipe()
        messageRouter.wipeOutbox()
        GossipMessageArchive.wipeDefault()
        StoreAndForwardMetrics.shared.reset()
@@ -556,6 +556,70 @@ struct CourierEndToEndTests {
        #expect(!stored)
    }

+    /// Regression for the relaunch amplification storm: redundant copies of
+    /// one message arrive as *distinct* envelopes (every seal uses a fresh
+    /// ephemeral key), so only the inner message ID can dedup them. The first
+    /// copy delivers; duplicates must stop right after the decrypt — no
+    /// second delivery, no second ack/handshake trigger downstream.
+    @Test func duplicateCourierCopiesDeliverInnerMessageOnce() async throws {
+        let alice = makeService()
+        let bob = makeService()
+        let bobDelegate = NoiseCaptureDelegate()
+        bob.delegate = bobDelegate
+
+        let bobKey = bob.noiseStaticPublicKeyData()
+        let first = try #require(alice.sealBridgeCourierEnvelope("once", messageID: "dup-1", recipientNoiseKey: bobKey))
+        let second = try #require(alice.sealBridgeCourierEnvelope("once", messageID: "dup-1", recipientNoiseKey: bobKey))
+        // Distinct seals: envelope-level dedup can never catch this pair.
+        #expect(first.ciphertext != second.ciphertext)
+
+        #expect(bob.openBridgedCourierEnvelope(first))
+        #expect(bob.openBridgedCourierEnvelope(second))
+
+        let delivered = await TestHelpers.waitUntil(
+            { !bobDelegate.snapshot().isEmpty },
+            timeout: TestConstants.defaultTimeout
+        )
+        #expect(delivered)
+        // Give a duplicate delivery a chance to surface, then confirm the
+        // second copy never reached the delegate.
+        let duplicated = await TestHelpers.waitUntil(
+            { bobDelegate.snapshot().count > 1 },
+            timeout: TestConstants.shortTimeout
+        )
+        #expect(!duplicated)
+        #expect(bobDelegate.snapshot().count == 1)
+    }
+
+    /// Acks for mail from absent senders (the usual couriered/bridged case)
+    /// queue for a future session instead of initiating a handshake
+    /// broadcast — otherwise every duplicate copy of every drop turns into a
+    /// mesh-wide handshake flood at an identity that cannot answer.
+    @Test func deliveryAckForAbsentPeerQueuesWithoutHandshake() async throws {
+        let ble = makeService()
+        let outbound = PacketTap()
+        ble._test_onOutboundPacket = outbound.record
+
+        // Nobody by this ID on the mesh (not connected, not reachable).
+        ble.sendDeliveryAck(for: "msg-1", to: PeerID(str: "00000000000000ee"))
+
+        let initiated = await TestHelpers.waitUntil(
+            { outbound.count(ofType: .noiseHandshake) > 0 },
+            timeout: TestConstants.shortTimeout
+        )
+        #expect(!initiated)
+
+        // Control: a peer that is actually around still gets the handshake.
+        let present = PeerID(str: "00000000000000ef")
+        ble._test_seedConnectedPeer(present, nickname: "present")
+        ble.sendDeliveryAck(for: "msg-2", to: present)
+        let initiatedForPresent = await TestHelpers.waitUntil(
+            { outbound.count(ofType: .noiseHandshake) > 0 },
+            timeout: TestConstants.defaultTimeout
+        )
+        #expect(initiatedForPresent)
+    }
+
    private func makeUnsignedAnnounce(from service: BLEService) throws -> BitchatPacket {
        let announcement = AnnouncementPacket(
            nickname: "Unsigned",
@@ -190,7 +190,11 @@ struct PrekeyEndToEndTests {
        #expect(message.content == "burn after reading")

        // 6. Redelivery tolerance: the same envelope arriving via another
-        //    packet (spray-and-wait) still opens inside the grace window.
+        //    packet (spray-and-wait) still decrypts inside the prekey grace
+        //    window (asserted at the crypto layer in NoisePrekeyTests), but
+        //    the duplicate is absorbed before delivery — the receiver dedups
+        //    on the inner message ID, so redundant courier copies never
+        //    re-deliver (or re-ack) the same message.
        let redelivery = BitchatPacket(
            type: MessageType.courierEnvelope.rawValue,
            senderID: Data(hexString: carol.myPeerID.id) ?? Data(),
@@ -203,9 +207,10 @@ struct PrekeyEndToEndTests {
        bob._test_handlePacket(redelivery, fromPeerID: carol.myPeerID)
        let redelivered = await TestHelpers.waitUntil(
            { bobDelegate.snapshot().count == 2 },
-            timeout: TestConstants.defaultTimeout
+            timeout: TestConstants.shortTimeout
        )
-        #expect(redelivered)
+        #expect(!redelivered)
+        #expect(bobDelegate.snapshot().count == 1)
    }

    @Test func withoutBundleSealingFallsBackToStatic() async throws {
@@ -35,8 +35,8 @@ struct BridgeCourierServiceTests {

        let service: BridgeCourierService

-        init() {
-            service = BridgeCourierService()
+        init(dedupStore: BridgeDropDedupStore? = nil) {
+            service = BridgeCourierService(dedupStore: dedupStore)
            service.bridgeEnabled = { [weak self] in self?.bridgeOn ?? false }
            service.relaysConnected = { [weak self] in self?.relaysConnected ?? false }
            service.publishEvent = { [weak self] event in self?.publishedEvents.append(event) }
@@ -175,6 +175,53 @@ struct BridgeCourierServiceTests {
        #expect(fixture.sealRequests.count == 1)
    }

+    @Test func publishedDropDedupSurvivesRelaunch() throws {
+        // Regression (field-verified amplification storm): the outbox that
+        // drives re-deposits is persisted, but the sender-side drop dedup was
+        // in-memory only — every relaunch republished the same undelivered
+        // message as a fresh drop, and relays hold each for 24h.
+        let fileURL = FileManager.default.temporaryDirectory
+            .appendingPathComponent("bridge-dedup-\(UUID().uuidString).json")
+        defer { try? FileManager.default.removeItem(at: fileURL) }
+        let recipientKey = Fixture.randomKey()
+        let messageID = UUID().uuidString
+
+        let fixture = Fixture(dedupStore: BridgeDropDedupStore(fileURL: fileURL))
+        fixture.sealResult = makeEnvelope(recipientKey: recipientKey)
+        #expect(fixture.service.depositDrop(content: "hello", messageID: messageID, recipientNoiseKey: recipientKey))
+        #expect(fixture.publishedEvents.count == 1)
+
+        // "Relaunch": a fresh service over the same store must refuse to
+        // publish the same message ID again (before even re-sealing it).
+        let relaunched = Fixture(dedupStore: BridgeDropDedupStore(fileURL: fileURL))
+        relaunched.sealResult = makeEnvelope(recipientKey: recipientKey)
+        #expect(!relaunched.service.depositDrop(content: "hello", messageID: messageID, recipientNoiseKey: recipientKey))
+        #expect(relaunched.publishedEvents.isEmpty)
+        #expect(relaunched.sealRequests.isEmpty)
+    }
+
+    @Test func seenDropEventDedupSurvivesRelaunch() throws {
+        // Same storm, gateway side: relays redeliver the whole 24h drop
+        // backlog on every launch; a relaunch must not re-open (and re-ack)
+        // events it already handled.
+        let fileURL = FileManager.default.temporaryDirectory
+            .appendingPathComponent("bridge-dedup-\(UUID().uuidString).json")
+        defer { try? FileManager.default.removeItem(at: fileURL) }
+
+        let fixture = Fixture(dedupStore: BridgeDropDedupStore(fileURL: fileURL))
+        let myKey = try #require(fixture.myKey)
+        fixture.service.refresh()
+        let event = try makeDropEvent(for: makeEnvelope(recipientKey: myKey))
+        fixture.service.handleDropEvent(event)
+        #expect(fixture.openedEnvelopes.count == 1)
+
+        let relaunched = Fixture(dedupStore: BridgeDropDedupStore(fileURL: fileURL))
+        relaunched.myKey = myKey
+        relaunched.service.refresh()
+        relaunched.service.handleDropEvent(event)
+        #expect(relaunched.openedEnvelopes.isEmpty)
+    }
+
    @Test func distinctDropsUseDistinctThrowawayKeys() {
        let fixture = Fixture()
        let keyA = Fixture.randomKey()
@@ -0,0 +1,118 @@
+//
+// BridgeDropDedupStoreTests.swift
+// bitchat
+//
+// This is free and unencumbered software released into the public domain.
+// For more information, see <https://unlicense.org>
+//
+
+import Foundation
+import Testing
+@testable import bitchat
+
+@Suite("Bridge drop dedup persistence")
+struct BridgeDropDedupStoreTests {
+
+    // MARK: - ExpiringIDSet
+
+    @Test func entriesExpireAfterLifetime() {
+        let start = Date(timeIntervalSince1970: 1_700_000_000)
+        var set = ExpiringIDSet(capacity: 8, lifetime: 60)
+
+        let inserted = set.insert("a", now: start)
+        #expect(inserted)
+        #expect(set.contains("a", now: start))
+        let duplicate = set.insert("a", now: start.addingTimeInterval(30))
+        #expect(!duplicate)
+
+        // Past the lifetime the slot is free again.
+        let later = start.addingTimeInterval(61)
+        #expect(!set.contains("a", now: later))
+        let reinserted = set.insert("a", now: later)
+        #expect(reinserted)
+    }
+
+    @Test func capacityEvictsOldestFirst() {
+        let start = Date(timeIntervalSince1970: 1_700_000_000)
+        var set = ExpiringIDSet(capacity: 2, lifetime: 3600)
+
+        set.insert("oldest", now: start)
+        set.insert("middle", now: start.addingTimeInterval(1))
+        set.insert("newest", now: start.addingTimeInterval(2))
+
+        let check = start.addingTimeInterval(3)
+        #expect(!set.contains("oldest", now: check))
+        #expect(set.contains("middle", now: check))
+        #expect(set.contains("newest", now: check))
+    }
+
+    @Test func removeReleasesSlot() {
+        let now = Date()
+        var set = ExpiringIDSet(capacity: 8, lifetime: 3600)
+        set.insert("a", now: now)
+        set.remove("a")
+        #expect(!set.contains("a", now: now))
+        let reinserted = set.insert("a", now: now)
+        #expect(reinserted)
+    }
+
+    @Test func initPrunesExpiredPersistedEntries() {
+        let now = Date()
+        let set = ExpiringIDSet(
+            capacity: 8,
+            lifetime: 3600,
+            entries: [
+                "stale": now.addingTimeInterval(-7200),
+                "fresh": now.addingTimeInterval(-60),
+            ],
+            now: now
+        )
+        #expect(!set.contains("stale", now: now))
+        #expect(set.contains("fresh", now: now))
+    }
+
+    // MARK: - Store round trip
+
+    @Test func snapshotRoundTripsThroughDisk() {
+        let fileURL = FileManager.default.temporaryDirectory
+            .appendingPathComponent("bridge-dedup-store-\(UUID().uuidString).json")
+        defer { try? FileManager.default.removeItem(at: fileURL) }
+        let recorded = Date(timeIntervalSince1970: 1_700_000_000)
+
+        let store = BridgeDropDedupStore(fileURL: fileURL)
+        store.save(BridgeDropDedupStore.Snapshot(
+            publishedDropKeys: ["msg-1": recorded],
+            seenDropEventIDs: ["event-1": recorded]
+        ))
+
+        let reloaded = BridgeDropDedupStore(fileURL: fileURL).load()
+        #expect(reloaded.publishedDropKeys["msg-1"] == recorded)
+        #expect(reloaded.seenDropEventIDs["event-1"] == recorded)
+    }
+
+    @Test func wipeRemovesTheRecord() {
+        let fileURL = FileManager.default.temporaryDirectory
+            .appendingPathComponent("bridge-dedup-store-\(UUID().uuidString).json")
+        defer { try? FileManager.default.removeItem(at: fileURL) }
+
+        let store = BridgeDropDedupStore(fileURL: fileURL)
+        store.save(BridgeDropDedupStore.Snapshot(
+            publishedDropKeys: ["msg-1": Date()],
+            seenDropEventIDs: [:]
+        ))
+        store.wipe()
+
+        let reloaded = BridgeDropDedupStore(fileURL: fileURL).load()
+        #expect(reloaded.publishedDropKeys.isEmpty)
+        #expect(reloaded.seenDropEventIDs.isEmpty)
+    }
+
+    @Test func nonPersistingStoreStaysEmpty() {
+        let store = BridgeDropDedupStore(persistsToDisk: false)
+        store.save(BridgeDropDedupStore.Snapshot(
+            publishedDropKeys: ["msg-1": Date()],
+            seenDropEventIDs: [:]
+        ))
+        #expect(store.load().publishedDropKeys.isEmpty)
+    }
+}
@@ -446,6 +446,34 @@ struct MessageRouterTests {
        #expect(transport.sentCourierMessages.count == 1)
    }

+    @Test @MainActor
+    func enqueueReplacementCarriesOverDepositedCourierKeys() async {
+        // Re-sending a queued message ID replaces the outbox entry; the
+        // replacement must inherit which couriers already carry the message,
+        // or the deposit retry re-burns the same courier slots (duplicate
+        // sealed copies to the same peer).
+        let recipient = PeerID(str: "00000000000000aa")
+        let recipientKey = Data(repeating: 0xBB, count: 32)
+        let courier = PeerID(str: "00000000000000cc")
+        let courierKey = Data(repeating: 0xCC, count: 32)
+
+        let transport = MockTransport()
+        transport.connectedPeers.insert(courier)
+        transport.updatePeerSnapshots([Self.snapshot(courier, key: courierKey, verified: true)])
+
+        let router = MessageRouter(
+            transports: [transport],
+            courierDirectory: Self.directory(recipient: recipient, recipientKey: recipientKey)
+        )
+        router.sendPrivate("Hello", to: recipient, recipientNickname: "Peer", messageID: "ck1")
+        #expect(transport.sentCourierMessages.count == 1)
+
+        // Same message ID re-sent (e.g. a resend while still queued): the
+        // courier already carrying it must not receive a second copy.
+        router.sendPrivate("Hello", to: recipient, recipientNickname: "Peer", messageID: "ck1")
+        #expect(transport.sentCourierMessages.count == 1)
+    }
+
    @Test @MainActor
    func courierBecameAvailable_ignoresTheRecipientThemselves() async {
        let recipient = PeerID(str: "00000000000000aa")