Fix: surface HTTP 413 and other 4xx errors via replication error event (issue #9165)

[TeapoyY]

Summary

Fixes #9165 — HTTP 413 during live replication is not surfaced via replication events.

Problem

When using PouchDB live replication with CouchDB, a \413 Content Too Large\ response from _bulk_docs\ is visible in DevTools but not exposed through any replication event. The \paused(err)\ event receives \err === undefined.

Root Cause

In \completeReplication(), only 'unauthorized'\ (401) and 'forbidden'\ (403) errors are emitted via the 'error'\ event. All other errors (including HTTP 413 Payload Too Large from _bulk_docs) fall through to the backoff/retry path, causing infinite retries without surfacing the error.

Fix

Treat 4xx client errors (except 401/403 already handled) the same as auth errors — emit them via the 'error'\ event and stop retrying. This is correct because 4xx errors indicate a problem with the request itself (e.g., content too large, bad request) that retrying cannot fix.

The fix is in \packages/node_modules/pouchdb-replication/src/replicate.js:

\\javascript
} else if (fatalError.status >= 400 && fatalError.status < 500) {
// Surface 4xx client errors (e.g. 413 Content Too Large from _bulk_docs)
// via the error event instead of retrying, since retrying won't fix them.
returnValue.emit('error', fatalError);
returnValue.removeAllListeners();
}
\\

Behavior Change

• Before: 413 errors during \�ulkDocs\ would silently retry forever; \paused(err)\ got \undefined\
• After: 413 (and other 4xx) errors are emitted via the 'error'\ event, so
  eplication.on('error', err => ...)\ receives the actual error with status/details

[janl]

Thanks for the contribution. Can you please use our pull request template?

[TeapoyY]

Closing - not a bounty issue