An MCP server that dropped before reconnect wiring finished now reconnects right away.
What's wrong with this entry?
If an MCP server's connection had already closed by the time the reconnect handler was attached, the reconnect is now scheduled immediately instead of waiting for a close event that will never arrive.
- The log line distinguishes a transport that closed later from one that "had already closed when wired".
- Reconnects now emit telemetry: success with the attempt count and whether it was found already closed, and failure as either
needs_authorexhaustedonce the attempt cap is reached.
had already closed when wired
Strings lifted out of the shipped bundle, so the claim above can be checked against them.