You can turn off the AWS machine and then do a snapshot. If anything at all goes wrong you can revert. For extra redundancy, you could do a pgdump to disk then upload to S3 which will take a while. But snapshots are generally considered reliable and have the advantage of being simple.
The threads about this issue have a number of sample logs. You would be looking for something like this. No confirmation is perfect but that would give you a pretty good idea of whether that's what you experience.
Alternately you can try the monit approach that @yanokwa suggested here. How frequently is this happening?