What
Memory recall is the step where Claude Code picks which of your saved memories are relevant to the current conversation. It uses a model to make that choice.
That step used to send a single request, and if it failed, the failure was only recorded. It now works through a list of candidate models. If one refuses the request, Claude Code logs a warning and tries the next. It records whether it had to fall back to another model, and separately records when every model on the list refused.
Why
Memory recall is less likely to fail quietly when a model refuses the request, so relevant memories are more likely to reach Claude.
Something disagreesSomething we can check disagrees with this entry, or the writer said they could not settle it.
The writer flagged doubt
Which models are on the candidate list is not stated.