Thanks for the reminder on this. I just posted not-so-brief release notes on HUMAnN 3.0.0 and the 201901b database update here:
Whether or not I would rerun a dataset would depend on the cost of the compute (which ends up being subjective). Personally, if I was working with a few hundred samples I would probably justify the rerun, but not for a few thousand samples. That might depend further on whether or not you had a species of interest among the ~600 new pangenomes added with this update.
Some of the other changes in this update (e.g. MetaCyc 24.0, the new UniRef mappings, changes to infer_taxonomy) could all be (re)computed quickly from gene family abundance profiles generated by the previous HUMAnN 3 – whether or not you want to repeat the gene family quantification would be the tougher question.