@Gargron Received spikes of requests up to 1,800/minute with 1 process x 6 threads of Puma in a container of 512 MB RAM + 512 MB swap. Surprisingly stable :)