Change Feed continuity after a failed upgrade to Cosmos DB API v2.0.0 Full Fidelity mode
25K reputation · 29 Nov 2025, 17:31 UTC
The goal is to understand whether Cosmos DB Change Feed guarantees exactly‑once processing of pre‑upgrade changes when a move to API v2.0.0 Full Fidelity mode fails and the account reverts to the legacy API.
Constraints include the lack of automatic checkpoint migration during version changes, the possibility that a processor already switched to Full Fidelity retains its configuration after rollback, and undocumented behavior for events occurring between the last legacy checkpoint and the failed upgrade.
Does the Change Feed processor automatically reconcile its checkpoints after a rollback to the legacy API?
What manual actions are required to prevent duplicate or missed events for changes made after the last legacy checkpoint but before the upgrade failure?
Is there a supported method to validate or migrate checkpoints when reverting to the previous API version?
cosmosdb