At 11:00 in the evening, there are two options for what they’re dealing with. Either:

1. They have made precisely no headway whatsoever in actually solving the bug, or
2. They have fixed the bug, but the debugging made them go “—wait, how did that ever work in the first place?!”

If it’s #1, odds are pretty good that there’s a random debug step they put in at 9:08 in the morning that’s screwing everything up now. If it’s #2, odds are pretty good that it actually didn’t work before, and now they’ve got to go back through the last six months of data and rectify it to fix that bug.