Routine shutdown no longer reports itself as an error

Every node in the portal runs small background watchers that keep it up to date — they notice when its code needs rebuilding, when a rebuild has been requested, and when its source files change. When the node itself shuts down, which happens routinely after a rebuild, after a period of inactivity, or whenever a page briefly opens a node to read its shape, those watchers naturally stop with it.

They were treating that as a failure. Each shutdown produced a burst of red "faulted — re-establishing" entries in the error log and armed a one-second retry against a node that no longer existed — steady noise that made the error log harder to read and kept a shutting-down node in memory a little longer than it needed to be.

A watcher now recognises its own node shutting down for what it is: the end of its job, not a fault. It stops quietly, nothing is retried, and the replacement watcher starts with the node's next activation. A watcher whose node is genuinely in trouble still reports it exactly as loudly as before.

Reconnecting…
The connection to the server was interrupted. Trying to restore it…
Trying again…
The connection could not be restored. Reloading the page…
The server was updated. Reloading the page to pick up the latest version.