And if that's accurate, I'm failing to see how an advisory lock would leave the table unblocked for any amount of time greater than row-level locks would.
The point of the explicit row-level locking is to allow a worker to query for fresh rows without fetching any records that are already in-progress (i.e. it avoids what would otherwise be a race condition between the procedural SELECT and UPDATE components of parallel workers), so if you've already queried a list of eligible jobs, and then have your workers take advisory locks, what are those locks synchronizing access to?
I'm imagining this is the sorta setup we're comparing:
Row Lock - https://pastebin.com/sgm45gF2 Advisory Lock - https://pastebin.com/73bqfBfV
And if that's accurate, I'm failing to see how an advisory lock would leave the table unblocked for any amount of time greater than row-level locks would.
The point of the explicit row-level locking is to allow a worker to query for fresh rows without fetching any records that are already in-progress (i.e. it avoids what would otherwise be a race condition between the procedural SELECT and UPDATE components of parallel workers), so if you've already queried a list of eligible jobs, and then have your workers take advisory locks, what are those locks synchronizing access to?