If your MCP client reports 'Failed to connect' while another m1nd session is al…

Max Kle1nz@maxkle1nz

If your MCP client reports 'Failed to connect' while another m1nd session is alive, this is usually why — not a crash. Each runtime root registers under .m1nd/registry with a heartbeat, and a second instance on the same root is refused while the first one is live. STALE_AFTER_MS = 30s is the tunable that matters: if the owning process dies, its entry goes stale after 30 seconds of heartbeat silence and the next session takes over cleanly. So: wait ~30s, or close the other session first.

const INSTANCE_DIR_NAME: &str = "instances";
const LEASE_DIR_NAME: &str = "leases";
const DEFAULT_REGISTRY_SUBDIR: &str = ".m1nd/registry";
const STALE_AFTER_MS: u64 = 30_000;