Bug: RelayPool.last_network_status race at queue/actor boundary

[alltheseas]

Problem

last_network_status in RelayPool is accessed from network monitor callback (on network_monitor_queue) and from @RelayPoolActor context without proper synchronization at the boundary.

Location

damus/Core/Nostr/RelayPool.swift lines 87-105

nonisolated init(ndb: Ndb?, keypair: Keypair? = nil) {
    network_monitor.pathUpdateHandler = { [weak self] path in
        Task { await self?.pathUpdateHandler(path: path) }
    }
    network_monitor.start(queue: network_monitor_queue)
}

private func pathUpdateHandler(path: NWPath) async {
    if (path.status == .satisfied) && self.last_network_status != path.status {  // RACE
        await self.connect_to_disconnected()
    }
    self.last_network_status = path.status  // RACE
}

Root Cause

The pathUpdateHandler closure runs on network_monitor_queue but accesses actor-isolated state. Multiple rapid network changes can create concurrent access.

Impact

• Network state tracking becomes unreliable
• Relays may fail to reconnect when network changes

Related Issues

• investigate: profile path update, dispatch queue #3432 (profile path update dispatch queue)

Suggested Fix

Make pathUpdateHandler explicitly @RelayPoolActor isolated, or use atomic storage for last_network_status.

Test Plan

• Test rapid network state changes
• Verify relay reconnection works correctly
• Run with ThreadSanitizer

Changelog-Fixed: Fix race condition in RelayPool network status tracking

[linear]

IOS-2283