Remote terminal daemon keeps running replaced binary after app upgrade

A desktop terminal app that uploads its own Linux helper binary to an SSH host and keeps a long-lived headless daemon there can hang on 'connecting' after an app upgrade. The new app overwrites the helper binary, but the old daemon keeps running the previous inode. On NFS homes, /proc/<pid>/exe points at a .nfsXXXX silly-rename file, which is how you spot it. New link processes from the new binary attach to the old daemon over its control socket and stall on the version skew. To diagnose, compare process start times with the binary's mtime and check /proc/<pid>/exe for .nfs or (deleted). The fix is to stop the stale daemon so the next connect spawns a fresh one; check first what child shells it hosts. A related symptom is that UI tied to a connected workspace, such as file-tree context-menu items, disappears, which looks like a lost feature but is just the unconnected state.