[joker-777]

Problem

unsafeResetDatabase() on the LokiJS adapter calls deleteDatabase(loki), which does loki.close(() => loki.deleteDatabase(...)).

close() only waits for a save if Loki is dirty. An autosave clears the dirty flags when it starts (in getLokiCopy), so while one is in flight Loki looks clean. close() then emits 'close' synchronously (Loki's listeners are synchronous), and IncrementalIndexedDBAdapter.deleteDatabase throws Error while deleting database - another operation is already in progress because operationInProgress is still true. The throw comes out of loki.close() inside the Promise executor, so the reset rejects.

Fix

After close(), wait for Loki's throttled save queue to drain (throttledSaveDrain) before calling deleteDatabase. Two small guards:

• throttledSaveDrain calls its callback twice when throttledSaves is disabled (a lokijs bug: missing return), so the drain is skipped in that case.
• The delete now runs in an async callback, where a throw would no longer reject the promise and would leave the reset hanging. It's wrapped in try/catch so it still rejects.

Note: when a save is pending, Loki's drain runs one more save before calling back. That save finishes before the delete starts, so it's harmless.

Test

New Jest test in src/adapters/lokijs/test.js. A fake incremental persistence adapter passed through _testLokiAdapter mimics IncrementalIndexedDBAdapter: it throws on save/delete while another operation is in progress and can keep a save in flight. The test starts a save, calls unsafeResetDatabase(), checks that nothing is deleted while the save runs, then finishes the save and checks that the reset completes and the data is gone. Without the fix it fails with "another operation is already in progress".

• jest src/adapters/lokijs src/adapters/__tests__: 140 passed
• eslint, tslint and flow focus-check on the changed file: clean

Independent of #1979 (persistence error reporting); the two touch lokiExtensions.js in different places and merge cleanly in either order.

🤖 Generated with Claude Code

[StasDoskalenko]

Hi @joker-777, thank you, nice find!

I maintain NitromelonDB, a maintained fork of WatermelonDB (New Architecture, Nitro Modules). It still ships the LokiJS adapter with the same deleteDatabase code, so this race exists there too, and a PR against our repo would be very welcome (src/adapters/lokijs/worker/lokiExtensions.ts). Our fork was rewritten in TypeScript (strict mode) to reduce the chance of bugs like this one slipping through, so I think the community would really benefit if you opened your PRs against it too.

Also possibly worth a look: on the web, our SQLiteAdapter now runs SQLite compiled to WASM (wa-sqlite) in a web worker, stored in IndexedDB. It's the same import and adapter as on native. There's no autosave queue to race with, since writes are real SQLite transactions, so it might be a better long-term option than LokiJS. More details are in the wa-sqlite adapter docs. One caveat: existing LokiJS data in the browser is not migrated automatically.

If you don't have the time or would rather not, no worries at all, I'm happy to port it myself. Just let me know. Thanks a lot!