WE are heavy users of AWS. SQS is the only service where we have had zero downtime. The only downside we have about SQS is you can pull out only 10 messages at a time (without batching). You can have parallel readers but they result in some duplicates. There is SQS FIFO but it is throttled.
It went down on us once for extended period in 2015. It was chaotic as you don't expect it to fail. If memory serves me right, even S3 suffered that day.
A whole slew of AWS services went down that day. Tim's not wrong when he indicates that almost every service in AWS has a dependency on it (Amazon has services split up in to tiers based on how much they can rely on other services for critical components, SQS is pretty high up in the tiering.)
I was on-call that day for an AWS service. There wasn't much I could do but sit muted on the conference call and watch some TV, waiting for the outage to be over.
