i guess it depends on the nature of the issue. like how would that happen? if the conclusion of the paper changed, in the worst case, i guess pursue some kind of formal retraction more likely it'd just be a bug in the code in which case i'd rerun experiments then update arxiv