feat: Propagate store errors and the availability check through the persistence wrapper

[joker23]

Summary

This is the wrapper-plumbing layer of the persistent store write-back recovery stack (follows #2009; the recovery engine itself lands in the next PR).

• PersistentDataStoreWrapper now forwards the underlying store's init and upsert errors to its callbacks instead of dropping them. A failed mirrored write is the signal the transactional store will use to detect an outage.
• The wrapper exposes isStoreAvailable only when the underlying PersistentDataStore implements it, so recovery logic can use the method's presence to decide whether it can poll for the store coming back.
• InMemoryFeatureStore.getAllRaw() returns a snapshot of the full data set, including deletion tombstones, for writing the in-memory state back into a recovered persistent store.
• The wrapper's init now populates the all-items cache with tombstones filtered out, matching the filtering the cache-miss path applies. Write-back init payloads can contain tombstones, and all() must never return them from the warm cache.

---

Note

Overview
Persistent store wrapper plumbing for write-back recovery: init and upsert now forward underlying store errors to their callbacks instead of swallowing them. On a failed init, the wrapper logs, clears item/all caches, and leaves _isInitialized false so reads and initialized() reflect persistence, not rejected in-memory data.

isStoreAvailable is exposed on the wrapper only when the core PersistentDataStore implements it, so recovery can detect poll support.

InMemoryFeatureStore.getAllRaw() adds an internal immutable snapshot of every kind, including deletion tombstones, for full-state write-back to a recovered store.

Cache fix on init: the warm all() cache is populated with tombstones filtered out, matching cache-miss behavior so all() never returns deleted items after init payloads that include tombstones.

Tests cover raw snapshots, cached all() with deletes, init/upsert error propagation, failed-init cache clearing, and isStoreAvailable forwarding.

^{Reviewed by Cursor Bugbot for commit 8b35e1e. Bugbot is set up for automated code reviews on this repo. Configure here.}

[github-actions]

@launchdarkly/js-sdk-common size report
This is the brotli compressed size of the ESM build.
Compressed size: 27065 bytes
Compressed size limit: 29000
Uncompressed size: 131676 bytes

[github-actions]

@launchdarkly/js-sdk-common size report
This is the brotli compressed size of the ESM build.
Compressed size: 26983 bytes
Compressed size limit: 29000
Uncompressed size: 131377 bytes

[github-actions]

@launchdarkly/js-client-sdk-common size report
This is the brotli compressed size of the ESM build.
Compressed size: 25437 bytes
Compressed size limit: 44000
Uncompressed size: 165420 bytes

[github-actions]

@launchdarkly/js-client-sdk size report
This is the brotli compressed size of the ESM build.
Compressed size: 32573 bytes
Compressed size limit: 34000
Uncompressed size: 116865 bytes

[joker23]

@cursor review

[cursor]

✅ Bugbot reviewed your changes and found no new issues!

Comment @cursor review or bugbot run to trigger another review on this PR

^{Reviewed by Cursor Bugbot for commit 10ebf5c. Configure here.}

[devin-ai-integration]

Devin Review found 1 new potential issue.

1 flag not posted on this PR by your GitHub settings — view it in Devin Review. (Configure)

[devin-ai-integration]

🟡 Recovery snapshot changes during serialization

When recovery serializes a segment, getAllRaw() shares its item object with the live store. replacer mutates that object, so the snapshot changes after capture.

Learn more

A raw snapshot copies each namespace map, but it retains references to the original flags and segments. Serializing a segment with generated target sets mutates its object in replacer. That changes both the supposedly fixed snapshot and the live in-memory store when the snapshot is passed to a persistence write. The same alias also lets any later in-place mutation of a stored item change the captured data.

Example: A segment has generated_includedSet = new Set(['alice']). After getAllRaw(), serializing that snapshot adds included: ['alice'] and deletes generated_includedSet on both the snapshot and the live segment; the captured object no longer has its original fields.

Recommended fix: Copy the item objects and nested mutable data when taking the snapshot, preserving Set values and other internal data types used by evaluation. Verify that passing the snapshot through sortDataSet does not mutate the live store or the captured data.

---

Was this helpful? React with 👍 or 👎 to provide feedback.

[joker23]

This does not affect correctness AFAIK