If a configured MCP server fails to spawn (e.g. ENOENT), the entire server is taken down as a result. This behavior is probably seen elsewhere child processes are used (procman, etc.). Instead, the error should be logged and a retry mechanism with a circuit breaker should be used.
If a configured MCP server fails to spawn (e.g. ENOENT), the entire server is taken down as a result. This behavior is probably seen elsewhere child processes are used (procman, etc.). Instead, the error should be logged and a retry mechanism with a circuit breaker should be used.