The health channel of one process: when it is due and its delivery. When it
last attempted an upload is the process's own fact, so a process created per
runtime layer keeps its own 240 s rhythm, and a process restored from a
checkpoint continues the checkpointed one's (lastAttemptAtMs,
continueProcess: a checkpoint carries it, as it carries the process timers).
The counters stay counted on the device until the upload is confirmed: a
failed or interrupted upload leaves them there.
The health channel of one process: when it is due and its delivery. When it last attempted an upload is the process's own fact, so a process created per runtime layer keeps its own 240 s rhythm, and a process restored from a checkpoint continues the checkpointed one's (
lastAttemptAtMs,continueProcess: a checkpoint carries it, as it carries the process timers). The counters stay counted on the device until the upload is confirmed: a failed or interrupted upload leaves them there.