Yep. But having said that, I'm not sure how useful it is in real life. For instance:
The server in the other room is a cheapie Gateway 6400 I bought for like $300 some years back when they needed to meet sales projections desperately. It uses a serverworks chipset, and requires ECC RAM. In the 2 years it's been running non-stop (I've been using it longer) the memory error light has come on once -- apparently cosmic rays conspired against me that day, and the ECC function corrected the issue.
I've run into client machines with freaky unexplained issues that pointed to RAM, but loading Linux on the system fixed the problem -- apparently (at least recent) Linux kernels notice when memory is misbehaving and map it out so it's not addressed.
So from limited personal experience, I'd think that freaky one-time-only memory issues happen very rarely (though you might want to pay for the ECC RAM anyway to avoid unexplained flakyness when it does), and my experience with Linux leads me to believe they do a good job even when running on known-bad memory.
I don't worry about it, and use whatever the server I like requires.
If you're just serving static web pages, don't worry. The worst that could happen is your machine crashes; a quick reboot and you're up and running again.
But with some applications - databases, complex physical modelling - a single error can spread and multiply and make a real mess of things. One error in the wrong place can corrupt an entire database index.
Random errors are pretty rare, and memory doesn't go bad very often. My day job is running large corporate databases, and there's no way I'd do that without ECC memory. (And mirrored disks, and offsite backups.) My current web server is ECC too, but the one before that wasn't, and it ran just fine for a year before I outgrew it.