[litbot-9000]

This branch belongs to the litbot-9000 bot. The yy prefix is only the allowlisted branch prefix; the branch is not yyforyongyu's.

Fixes the unit-race tags="test_db_sqlite" failure reported in #11507.

Problem

The first nightly run (job 127023741626) failed with:

--- FAIL: TestLightningWallet/btcwallet/neutrino:get_transaction_details (1.66s)
    test_interface.go:1791: expected transaction fe2b4114… to be confirmed

sendCoins mines the transaction and then calls waitForWalletSync on both wallets. That waits until the wallet's best block equals the miner's tip and IsSynced() returns true. With neutrino, though, the wallet's sync height can advance before the matching block filter has been processed and the transaction's confirmation recorded, because that happens asynchronously. assertTxInWallet then reads the transaction list exactly once, and in that window it sees the transaction as still unconfirmed.

Fix

assertTxInWallet now polls with wait.NoError for up to 30 seconds until the transaction appears with the expected confirmation status, instead of failing on the first read. The final check is the same. A wallet that never records the confirmation, or that records one that's wrong, still fails, with the same message. Every caller benefits, including the assertions that expect a transaction to stay unconfirmed: polling can't make those pass spuriously, because they already require the transaction to be found in the unconfirmed state.

Verification

• go vet ./lnwallet/test/... and gofmt are clean.
• I ran the neutrino transaction_details, transaction_details_offset and get_transaction_details subtests locally against a btcd harness: pass.
• I ran the full lnwallet/test/neutrino and lnwallet/test/btcd suites locally with -race: pass (neutrino 201s, btcd 108s).

The race itself is timing dependent, so it doesn't reproduce on demand; the nightly run is the real check.

Not covered

The other failure in #11507, Run windows itest (coop_close_exceeds_max_fee and wallet_import_pubkey/BIP-0049, both timeouts), is unrelated. Those are timeouts on slow Windows runners. The same job also failed in a merge-queue run on 2026-11-03. It's not a required check, and it isn't changed here.

[github-actions]

🟢 PR Severity: LOW

gh pr view | 2 files | 65 lines changed

🟢 Low (2 files)

• lnwallet/test/test_interface.go - test-only change: makes assertTxInWallet poll for the expected confirmation status instead of reading once, fixing a flaky unit-race failure; no production wallet logic touched
• docs/release-notes/unreleased/11508.md - release notes entry

Analysis

Although lnwallet/* is normally CRITICAL, the only non-docs file here lives in lnwallet/test/, a dedicated test-harness package (analogous to lntest/) used exclusively by interface tests across wallet backends. The diff only changes polling/retry behavior in a test assertion helper to fix a timing-dependent flake reported in #‌11507 — it does not modify any production wallet, signing, or funding code. Combined with the release-notes entry, this PR is test-only and best-effort review is sufficient.

---

_{To override, add a severity-override-{critical,high,medium,low} label.}