JeffAbrahamson
commented
1 year ago I increasingly think this is a utf-8 decoding problem on the part of some email recipient, although that remains to be proved. In any case, bad tokens appear not to be in our remit: what we emit is good (we have tested), and so when we get bad ones, we should log them and ignore.
Log them, do something reasonable if interactive, but don't emit a stack crawl. Just log what we recieved, why we couldn't handle it, and what we did.
This happens here:
/ml/manage/press/<token>/