Created in response to JADE Contact #68211. Details from this Contact are:
One of our customers experienced an issue yesterday with their production RPS replication, which was triggered in the first instance by an SQL timeout. It appears that the RPS tracking halted (correctly) due to this issue, but then in the first (of 10) automatic retries, a 3211 (SDS tracker thread is busy) exception occurred, causing the datapump to stop, and the automatic retry process was abandoned. The documentation of the 3211 exception says "Wait until the tracker process has completely started or stopped and then retry the operation if it is still required", so we would expect the automatic restart of the RPS datapump to keep retrying, at least for the configured 10 retry attempts.
Manual restart of the datapump some time later worked OK. It would be much better if the automatic restart/retry handled this situation in a more robust manner.