Reproducible crashes marked as non-reproducible?

I'm seeing the second "non-reproducible" crash in the last couple of weeks which turned out to be reproducible: https://oss-fuzz.com/testcase-detail/5227577549848576. Both issues were caused by assertion failures aborting the fuzzer and not by sanitizers. I also can't see any obvious source of non-determinism but maybe I'm missing something.

Start with the linked OSS-Fuzz testcase 5227577549848576 and investigate how assertion-failure crashes are classified as non-reproducible. Compare that path with sanitizer-triggered crashes; done means reproducible assertion crashes are classified consistently.

Written by the indexing model from the issue text.