A save caught by the owner restarting lands again instead of failing

When you save content that lives on another hub and that hub happens to be shutting down at the same moment — a recycle, a restart, a whole-portal stop — the owner tells your side "I did not apply this; retry against my successor", and your side retries automatically. That has worked for a while.

Yesterday's fix for saves that hung when the owner stopped watching introduced a second place that could give that same "retry" answer — and it gave it one step too early. The owner's stream ends partway through its shutdown, while the old hub still holds the address. Answering at that instant sent the automatic retry straight into the hub that was still going down, which refused it, and the save failed with an unexplained error instead of landing. During a full shutdown the early answer was worse: it was sent over a route that no longer delivers, and because a save is answered only once, the proper answer — the one that does reach the waiting save — was never given. The save then waited out its whole confirmation window in silence.

The early answer now stands aside whenever the owner is shutting down. The verdict comes, as before, from the owner's final shutdown step — the point at which it has released the address — so the retry lands on the fresh hub and the save completes, and it is handed to the waiting save directly, so a portal-wide shutdown no longer leaves a save hanging. The "owner stopped watching" case keeps its prompt answer; the two are told apart by whether the owner is going down.