[Bug]: VSS ChannelMonitor desync causes unrecoverable "LDK Build error: Read failed" on wallet restore

[piotr-iohk]

What happened?

After restoring a mainnet wallet from mnemonic, the LDK node fails to build with:

LDK Build error: Read failed.

The underlying cause is a stale ChannelMonitor in VSS. During node build, LDK loads channel afa669433240654dc0f30b01a8be73cf5fa9ec8446d6c0d6c2ec77a51a3e8b4a with ChannelManager at update_id 1344 but ChannelMonitor at update_id 1162 (182 updates behind). LDK refuses to start as a safety measure to prevent fund loss.

A ChannelMonitor is stale compared to the current ChannelManager!
This indicates a potentially-critical violation of the chain::Watch API!
The ChannelMonitor for channel afa669...8b4a is at update_id 1162 with update_id through 1162 in-flight
but the ChannelManager is at update_id 1344.

The node enters ErrorStarting state and every restart attempt fails identically since the same corrupted data is read from VSS each time.

The same failure also reproduced on iOS when restoring the same wallet.

Expected behavior

Wallet restore from mnemonic should rebuild the LDK node from VSS-backed state and start successfully. ChannelMonitor and ChannelManager update_ids in VSS should always be in sync.

Steps to Reproduce

1. Have a mainnet wallet that has been actively sending Lightning payments over time (this wallet was used for nightly E2E tests sending 5 sats to Strike via LN address)
2. Wipe the app and restore the wallet from its mnemonic
3. App restores app-level data from VSS successfully (METADATA, SETTINGS, WIDGETS, BLOCKTANK, ACTIVITY)
4. LDK node build fails with "LDK Build error: Read failed" due to stale ChannelMonitor in VSS
5. All subsequent restart attempts fail with the same error

Logs / Screenshots / Recordings

See attached bitkit_logs_1773391029542.zip.

⚠️ Note that only entries since the last wallet restore are relevant.

Key log timestamps (from bitkit_2026-03-13_05-14-30.log):

• 08:32:59 — Wallet wipe initiated
• 08:33:01 — Node stopped, LDK storage wiped
• 08:33:18 — Mnemonic entered
• 08:33:19 — VSS client connects with store id bitkit_v1_bitcoin_35c208c617a401dd2c66260378796ecb621e
• 08:33:29 — Full app-level VSS restore succeeds
• 08:33:30 — LDK node build starts, reads channel state from VSS
• 08:33:32 — ChannelMonitor stale error — monitor at update_id 1162 vs manager at 1344
• 08:33:32 → 08:36:57 — 5 repeated build attempts, all fail identically

Bitkit Version

Likely corrupted in v2.1.0 (last successful Strike payment was on 11 Mar with v2.0.3; first failure on 12 Mar). However, a similar wallet sending to WoS on the same builds is not corrupted, so the trigger may be timing/usage-dependent rather than purely version-based.

Device / OS

Android (also reproduced on iOS)

Reproducibility

Always

Additional context

• Environment: Mainnet
• Wallet history: Restored from backup. The wallet had been actively used for nightly E2E Lightning payment tests (sending 5 sats to Strike LN address).
• Root cause hypothesis: During normal wallet operation, ChannelMonitor updates for channel afa669... were not being durably persisted to VSS even though ChannelManager updates were. Over time this accumulated a 182-update gap. The wallet worked fine on the original device (local state was consistent), but the VSS backup became silently corrupted. The corruption is only exposed when restoring to a fresh device that relies entirely on VSS state.
• Impact: Wallet is permanently unrecoverable via normal restore flow. Lightning channel funds are inaccessible until the counterparty (Blocktank 03816141...) force-closes the channel.
• Cross-platform: Same failure on both Android and iOS restore, confirming the issue is in the VSS-persisted LDK state, not platform-specific.