MCP retries are abandoned if you switch accounts during the backoff wait.
What's wrong with this entry?
The retry loop for remote MCP servers that failed transiently now snapshots the signed-in account before backing off and abandons the queued retries if it changed, because those rows belong to the previous account. This applies on every retry, with no flag.
- The loop only re-attempts servers whose remembered connect result was actually discarded, and skips the round trip entirely when none were.
- The reconnect path captures the account before its retry timer and returns early if it no longer matches.
- The connector-connect path likewise discards results superseded by a newer attempt.
[MCP] Retry: identity changed during backoff, stopping (these rows belong to the previous account)
Strings lifted out of the shipped bundle, so the claim above can be checked against them.