I was under the impression that AWS gives no SLAs on an individual instance level or even an AZ level, but a region should always remain up (obviously they're not 100% at that), therefore you need to thoroughly ensure your config is multi-AZ and pay AWS accordingly.
As far as actual observed reliability, I'm curious to hear any others' experiences, but that seems pretty anecdotal.
As far as actual observed reliability, I'm curious to hear any others' experiences, but that seems pretty anecdotal.