You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
{{ message }}
Repository navigation
Session replay: early chunks saved under the wrong session #537
If you send events from another pipeline with a stable __deviceId, for example to match session recordings to a session tracked through a CDP, the first replay chunks end up under a different session from the events
Replay chunks recorded before the first track response have no sessionId, so the SDK queues them. When the session arrives, flush() sends them as they are (buildFlushPayload returns replay payloads unchanged), and the server falls back to the session it works out from IP and user agent. Usually that's the same session. With __deviceId set, it isn't. The early chunks, including the full snapshot, are stored under a second session, and the replay of the real session starts without a snapshot
Repro: enable session replay, send a track with properties.__deviceId set to a stable id, then check session_replay_chunks. The early chunks have a different session_id from the event
changed the title [-]Session replay: chunks recorded before the session opens are filed under a different session[/-][+]Session replay: early chunks saved under the wrong session[/+]on Oct 9, 2026
Activity
[-]Session replay: chunks recorded before the session opens are filed under a different session[/-][+]Session replay: early chunks saved under the wrong session[/+]mrfambo commentedon Oct 18, 2026
Use
EXPERIMENTAL_PROFILE_BACKFILL: "1"fix(sdk): bind early session replay chunks to the tracked session (Op…