fix: auto-recover stale network graph cache

[jvsena42]

This PR adds automatic validation and recovery for stale RGS network graph caches that cause Lightning payment routing failures.

Description

When the app uses RGS (Rapid Gossip Sync) delta updates, nodes that are missing from a stale cached graph won't be restored by incremental syncs. This caused RouteNotFound errors when trying to pay certain destinations like Blink wallet.

1. Adds validateNetworkGraph() to check if trusted peers (Blocktank LSP nodes) are present in the graph
2. Adds resetNetworkGraph() to delete the cached graph file when stale
3. Validates the graph automatically on node startup and triggers a fresh full sync if trusted peers are missing
4. Skips validation when graph is empty to avoid false positives on first launch

Preview

pay-to-blink.mp4

QA Notes

1. Verify stale graph auto-recovery

1. Launch app with existing wallet that has a stale network graph cache
2. Check logs for: Network graph missing X trusted peers
3. Verify logs show: Network graph is stale, resetting and restarting...
4. After restart, verify: Network graph validated: all X trusted peers present

2. Verify normal startup (no stale cache)

1. Launch app with valid network graph
2. Verify logs show: Network graph validated: all X trusted peers present
3. No restart should occur

3. Verify payment routing works

1. After graph validation passes, attempt a Lightning payment to Blink wallet
2. Payment should route successfully without RouteNotFound error

[claude]

Code review findings

[jvsena42]

Note: This is more like a fallback. The source issue must be investigated further

[jvsena42]

Testing...

[jvsena42]

Relevant Logs

2026-02-05 10:43:07.444  5592-7341  LDK                     to.bitkit                            I  2026-02-05 13:43:07.444 INFO    [ldk_node::builder:1554]             Migrated network graph from VSS to local cache (RGS timestamp: 1770282000)
2026-02-05 10:43:13.432  5592-6221  APP                     to.bitkit                            D  2026-02-05 13:43:13.431 DEBUG   [LightningService.kt:310]            Network graph validated: 3/3 trusted peers present - LightningService

I had to uninstall the previous app because of different signatures
These logs are after reinstall and restore the Wallet from mnemonic

[jvsena42]

Succeeded on multiple attempts of 4k sats

[pwltr]

Concept ACK

[jvsena42]

related LDK PR
lightningdevkit/ldk-node#692

[ovitrif]

tAck

Test results:

1. Verify stale graph auto-recovery: 🟡 norepro
2. Verify normal startup (no stale cache): 🟢 success
3. Verify payment routing works by paying to Blink: 🔴 error

[ovitrif]

idea: consider distinguishing "fresh install / post-reset" from "corrupted graph" here.

Right now an empty graph passes validation unconditionally. After a graph reset + restart, this is expected — RGS hasn't synced yet. But if a graph has a non-null latestRgsSnapshotTimestamp and is still empty, that could indicate corruption rather than a clean slate.

A possible approach:

if (graphNodes.isEmpty()) {
    val rgsTimestamp = node.status().latestRgsSnapshotTimestamp
    if (rgsTimestamp != null) {
        Logger.warn("Network graph is empty despite RGS timestamp $rgsTimestamp", context = TAG)
        return false
    }
    Logger.debug("Network graph is empty (fresh install), skipping validation", context = TAG)
    return true
}

This would catch the edge case where VSS migration or file corruption produces a graph file that RGS treats as "up to date" but contains no nodes — which is exactly the failure mode we saw on mainnet (0 RGS updates applied on top of a stale VSS snapshot).

Not a blocker — just a defense-in-depth idea for a follow-up.