This handles an error that's very unlikely to happen. I think that today it's the first time I've seen this.
We were not handling TransactionTimestampNotWithinRange in notify jobs. At first glance, this doesn't look necessary because the timestamps in those jobs come from the listener processes and not the user. However, there are a few corner cases that make this necessary. Please check the explanation I added here.
We need to handle TransactionTimestampNotWithinRange, otherwise, notify jobs are re-enqueued even though they can't be processed correctly.
This handles an error that's very unlikely to happen. I think that today it's the first time I've seen this.
We were not handling
TransactionTimestampNotWithinRange
in notify jobs. At first glance, this doesn't look necessary because the timestamps in those jobs come from the listener processes and not the user. However, there are a few corner cases that make this necessary. Please check the explanation I added here.We need to handle
TransactionTimestampNotWithinRange
, otherwise, notify jobs are re-enqueued even though they can't be processed correctly.