There are a lot of special cases/retry logic for distributed transactions, but some code paths are overloaded, so its hard to determine whether we are in a special case or not if we don't have a repro. I think it could make debugging HELP/BFs more straightforward if we added more logging to indicate we are in these special cases or include more information in the existing logs.

Specific places that could be helpful:

• If we are reusing the same txnNumber
• If we're retrying on a transient error, we should log the error because right now there is no error log, we just see two consecutive "New transaction started" logs for the same number
• Log the pending participants we're clearing when clearing pending participants. (We log when we add them and include the id, so we should log the id in the clearing function too, especially because there is a case where we clear the participant list but do not abort them)

This came out of BF-40166, so could be helpful to look at the comments there for other places to add logs.