SRV covers the largest audience -- people with an email address and the
ability to resolve a domain name. Let's start there.
SRV doesn't solve the problem for nomadic users on slow links who don't
want to wait forever while their MUA submits mail to the ISP associated
with their email address.
SRV doesn't solve the problem for users on networks for which one of
the protocols is blocked.
SRV doesn't solve the problem for users whose mail is forwarded from
the From address to some other address or message store.
It took me about 3 minutes to think of these cases. There are probably
Taken together, those cases might or might not add up to a situation
where SRV is not good enough - doesn't cover a wide enough set of cases
- would invite competing solutions which would make configuration more
rather than less complex. The only way to have any confidence in such
a design choice is to do some case analysis.
Does a bridge stay up because it looks pretty? Or because the market
likes it? Or because someone took the trouble to estimate the loads
that it would be subjected to (allowing for some margin of error and
projecting for increased use over time) and to analyze the ability of
the structure, as designed, to handle those loads?