I only use a decay function to see how "hot" a chunk is - not for forgetting old ones. What concerns me more are memory chunks with errors in them - they need to be corrected/removed by some other mechanism, not by decay (since they might get retrieved often).

Fair concern, and a real limitation. Decay alone won't catch a wrong memory that gets retrieved often, recall reinforces it instead.

The current answer, contradiction detection on store. When a new fact contradicts an existing one, the older entry gets superseded via update_memory. Where this does work well, configuration drift. If your old OAuth config is stored and a new one comes in that's semantically similar but factually different, similarity is high enough to trigger an update, the stale one falls out. Subtle errors that don't contradict anything new still slip through.

Active validation and periodic recheck are the better long term fixes. Not there yet.