It's likely that the number of requests to this instance was in excess of 2 * n_cpu (which I've seen given as advice). I am currently scaling based on CPU utilization rather than number of connections. Let me know if you think this is unwise / could be the cause.
As an experiment, I have built the imagor Docker container on the
The imagor service has been crashing periodically with the following error message:
A couple of relevant notes: