It's a bit weird, but I think it's meant so a user can be created from some external auth source without an e-mail address, but then they're forced to set one on login. However this currently also applies to the admin if they edit the user again, so perhaps it should still be optional if an admin is saving the user? Or it just needs redesigning.

If nothing depends on it, I'd fully remove the validator. I guess it won't wreck anything since there already could be users without email.

The only thing that would use it are the e-mails on failed reports (to the owner), so check there's a fallback or at least graceful exit in these cases. We could keep the validator, but permit an empty e-mail address too.

I have entering an email address everywhere - e.g. I am evaluating a software, I know I will not be sending any reports and I need to enter it. I vote for optional e-mail address and changing our sender helper. If the e-mail provided was nil, let's just put warning in logs and skip the send.

+1 for optional

This was discussed again recently https://community.theforeman.org/t/kerberos-cross-domain-signin-fails-foreman-webui/27460/16 specifically this comment https://community.theforeman.org/t/kerberos-cross-domain-signin-fails-foreman-webui/27460/4