We capture app_instance_id from the Firebase SDK during user login/signup and store it in our backend. All journey events are sent from the backend via Measurement Protocol using this stored app_instance_id.
We observed that session_start and other events fired directly from the Firebase SDK show 100% app_version coverage for the same users. However, for events sent via Measurement Protocol using the same app_instance_id, only ~50% of events get app_version tagged correctly — the rest show “not set”.
Since the app_instance_id is captured only at login/signup and reused for all subsequent backend events, could this be the reason? Or is there a known limitation where Measurement Protocol events don’t reliably inherit app_version from the SDK session even when a valid app_instance_id is provided?
Has anyone faced this issue? Any insights on how Firebase handles app_version inference for Measurement Protocol events vs SDK events?