However, the SLAs that I can see are 'initial response', which for better or worse, sometimes happen. However, the most egregious example of how broken that model is, we had a business critical system down ticket down that got an answer of 'looking into it' in 19 minutes, then ~1:45 an answer of "oh, yeah, I checked and this and this DB you talked about didn't reboot", which was true, the person that made the ticket pointed out the wrong instance. 3 hours after I point the correct DB we get told, yes, you are right, it rebooted. And then on the 1st of April I ask for a timeline of how the writer/reader nodes changed during the reboot, I get an answer on the 20th, even though we pushed in the ticket and through our TAM for an answer.
Generally, I'm not really sure how people are so happy with AWS support, every issue that is even the least bit complicated gets wrong answers (took a month for us to convince people, though meetings with our TAM, that it's not a configuration or capacity issue that's causing some issues) or just sits there for ages. Once we got an engineer on the Aurora team assigned to the case I used as an example, we started making progress, we even got a patched version of PostgreSQL RDS deployed, but once the bug was identified, we had to work around it, since a fixed version was not an option in less than a few months.
All in all, for the hundreds of thousands of dollars we're spending a month on support, I have a feeling that when we need it, we're kind of alone.