* test(auth): fix flakiness in browser integration and webdriver tests
* Fix WebDriver email sign-in test flakiness by using robust JS click/fill
* Ensure deleteApp is called even if resetEmulator fails
* test(auth): catch emulator and deleteApp cleanup errors to prevent failing afterEach hook
* build: do not throw on missing deploying token if deployment is not needed
* build: add --include-dependencies to build_changed so dependents compile successfully
* build: workaround glob-stream singular glob Node 22 bug in firestore integration tests
* chore: revert allowEmpty: true
* Fix querystring to always use apiKey even if a key is specified in the request
* Update packages/auth/src/api/index.ts
* Fix test helper URL generation to match new querystring order
* Run yarn format
---------
Co-authored-by: Hanzo Dev <[email protected]>
* fix(auth): retry IndexedDB availability check on startup failure
In environments like Chrome Extension Manifest V3 Service Workers, the initial IndexedDB transaction upon a cold start can sometimes drop or timeout. Previously, `_isAvailable()` would immediately catch the error, assume IndexedDB is broken, and silently fall back to `inMemoryPersistence`.
This updates `_isAvailable()` to use the `_withRetries()` helper, ensuring it benefits from the same retry logic as actual storage operations. This prevents the SDK from incorrectly abandoning existing IndexedDB auth state.
Fixes#9732
* style(auth): run yarn format
* chore: add changeset for IndexedDB availability check retry
* fix(auth): robust IndexedDB connection handling and test stability
