It is a known issue and devs and top witnesses are looking into it. The fix will be issued at some point soon (hopefully).

From what I know is that there is a problem with scheduler where it "randomly" picks one backup witness and stuck it into a production loop, while everyone else is almost being ignored. It is not dependent on the witness rank (we see witnesses from rank (20-30) to a rare case of rank 100-ish being stuck in a loop, and doesn't pose any security risk.

Good eye for spotting this one!