Today’s entirely foreseeable headache in the offing: https://github.com/pandas-dev/pandas/issues/37654#issuecomment-845063168
I’m fairly sure that forthcoming changes to dateutil will break any pandas code using dateutil zones relatively soon.
Even if it gets fixed in pandas today I’m sure it will be annoying.
Not that there’s much action you can take as an end user of pandas other than getting your deployment pipeline in good enough shape that you can deploy the latest version of pandas on short notice.
If the underlying problem isn’t fixed, the real problem is going to be when pytz transitions to using zoneinfo under the hood: https://github.com/stub42/pytz/issues/48#issuecomment-720235305
Exacerbated by the fact that pytz bundles in the time zone data, so you may be stuck on old tzdata if you pin pytz!