MCP server reconnect failure gives no diagnostic information

Problem

When an MCP server fails to connect or reconnect (via /mcp), the only feedback is:

Failed to reconnect to <server-name>.

There is no additional information about why it failed — no stderr output, no exit code, no timeout indication, nothing. This makes debugging MCP server startup issues extremely difficult.

Expected behavior

The error message should include at least:

• The command that was attempted
• stderr output from the failed process (if any)
• Whether it was a timeout vs crash vs connection refused
• The exit code if the process exited

For example:

Failed to reconnect to hn.
  Command: bash -c "exec clojure -M -m hn.mcp"
  Exit code: 1
  stderr: Exception in thread "main" java.io.FileNotFoundException: ...

Context

I'm building custom MCP servers (Clojure stdio servers). When they fail to start, the only way to debug is to manually run the command in a terminal and see what happens. The /mcp dialog gives zero diagnostic info.

claude mcp get <name> sometimes reports ✓ Connected even when /mcp in-session fails to reconnect, which adds to the confusion.

Environment

• Claude Code on Linux (WSL2)
• MCP servers configured via claude mcp add-json --scope user
• stdio transport