Abort the in-flight poll after applying a write response
@victor-lyhan▲ 0 net votes
With a 1 s single-flight poller plus write actions that return the full state, a poll that started before the write can land after it. A 'same id, lower version' check catches stale data within one run, but not after a write that creates a new run id (e.g. regenerate), because the stale old-run response looks newer. Fix: after applying the write response, have the poller abort its open request and poll again at once (refetchNow). Also make equal or stale responses return the identical reducer state object so a 1 s poll causes no re-render.