>
But it really does sound like you're going to need to roll your own service based on your major complaints.Ideally that'd be more the norm, but unfortunately it's significantly more hassle than would be ideal. Uptime is an issue, though most senders should retry through brief outages. Management/security updates of course aren't negligible, though in principle fairly manageable. Probably the big issue is deliverability to the big players like Gmail. They're as bad at support and opaque there as everywhere else, if you don't know someone and/or aren't big enough regardless of what boxes you check you can have email just vanish into the ether which is generally unacceptable.
One solution might be to use a decent commercial SMTP relay service to proxy that part (ie., Amazon SES/Mailgun/Postmark/SendGrid etc). Those tend to be aimed at commercial operations doing high volume but can start low enough to be worth considering. But all in all it's hard to do well despite email being so old and undemanding anything made in decades can run it fine. The foundations just aren't there to cheaply autoscale around abuse. Like, ideally we'd have secure DNS to depend on, that in turn would allow standards for just having a CA root cert for a domain put right into the domain's DNS records, in turn allowing distributed signing/encryption and making spoofing of email addresses non-trivial. No sending of any kind allowed without having it signed by something in the domain CA's trust chain. Remaining abuse after that could managed far more easily, reliably, and perhaps most important deterministically.
No such luck though! Leaving us stuck paying ludicrous amounts and dealing with a lot of crap for a service that an RPi is already has orders of magnitude more power than it should need for :(.