Glenn has identified the real cause. At the stage where I was looking closely at WeBWorK code revision, I was only looking at PGchoicemacros.pl. But it turned out to be a change in PGbasicmacros.pl. In any case it is an issue with perl randomizing hash order, and it sounds like the PR I opened (https://github.com/openwebwork/pg/pull/1102) will address this. If you can wait for a parallel PR to be merged into the main pg branch, that is best. If you cannot wait, you could manually apply those changes.
WeBWorK Problems
Unrandomizing answers in pgchoicemacros.pl
This forum has a limit to the number of forum postings you can make in a given time period - this is currently set at 10 posting(s) in 1 day