See PHI765. If you write a rule like this which has both a "block" action and a "send an email" action:
[ Author ] [ is ] [ Weasel ]
[ Block push with message ] [ No weasels allowed! ]
[ Send an email to ] [ email@example.com ]
...the "send an email" action will never have any effect. When we block a push, we don't send email about it.
The intent of the "send an email" action is almost certainly "send an email when a push is blocked".
I think this is a generally reasonable thing to want.
In contrast, the intent with this rule:
[ Always ]
[ Send an email to ] [ firstname.lastname@example.org ]
...is almost certainly "send an email when a push is not blocked".
We could reasonably infer these intents, although we'd have to make some code changes to how "Send me an email" and "Block" work and manage data.
I don't love this because giving Herald rule actions contextual magic seems generally spooky, but I think we can pretty reasonably infer the intent in both cases.