Yeah, I have to fix that 1.28.
Seems okay. Although I only gave it a cursory look, it's a new package and does not seem to conflict with anything.
Of course now instead of 23 dependencies it pulls 35 dependencies but oh well.
This works - resolves problem and cross-checks with a few other PDFs (pulled directly from Koji due to QA failure).
I verified that fix for 1065565 and found no other regressions.
Tested python3-eventlet-0.30.1-1.fc33 and it works.