Yeah, it only updates at 100%.
As for a monitoring use case, although we can estimate the time
remaining based on the change in items processed, that sort of thing
might not be as useful in cases where errors are occurring or
monitoring is stopped and then resumed at a future time (i.e. loss of
previous state information). I guess I expected that the
processingTimeMillis would be a running total provided by the API.
Generally speaking, I think it would give the receiving side of the
API some insight into what is happening throughout a job's life,
especially for stalled jobs and error cases where it's often difficult
to know what exactly is going on with the API side.