That's not unusual. If you read lots of launch-day CPU reviews, they often note issues they encountered. Many times, this is due to using pre-release BIOS and sometimes Intel/AMD even manage to turn around a fix in time for the review's completion.
These fixes took like 2 months, though. Stuff definitely wasn't done in time for launch that should've been. I have to wonder whether that was due to cancelling the 20A node, which might've meant they had far less time than usual with silicon in house, prior to launch.
It's probably a good thing the socket wasn't new and that they reused Meteor Lake's I/O tile. There are some Meteor Lake-S engineering samples floating around, so I'm sure those were used to develop & debug basic platform support, before Arrow Lake silicon ever arrived.