- Latency of the oregon region is very high. A simple health endpoint is around 500ms from north america and europe, 1.3s from Asia/Australia (we have been monitoring it for many months with better uptime)
- Services tend to die at least once a week, so you do need at least two instances all the time
- Logs are often delayed for a long time, though it has been more severe recently.
- Build time is just horrible, our rust service takes around 40m to deploy. I get that it is free, but at least give me a docker registry so I can build it elsewhere and push the image
- No option to pay for support 24/7, so if there is an outage during the US night you are on your own