You are right, there is a bug that is corrupting images. I still haven’t figured it out what causes that bug. Funnily enough that only happens on the production infrastructure meaning it’s the server itself. My guess (and I think I am right with it) is that it takes longer to upload the image to the S3 (it remains on the server for a few seconds for upload to save space) than actually having the image on the S3 creating some sort of blackhole for images depending on the current load. The only way to fix this is to get a fiber connection to the server (which is currently in the works, but I can’t tell when exactly it will happen). I will talk about it with Zack and get back if we know more about it.
tl;dr it’s a known bug, but we’re not sure how to fix it in our current situation