It goes without saying that hosting MESA entirely on the cloud would be... expensive.
In order to ensure the stream is sustainable from a cost standpoint, I optimized early on and built up a homelab setup.
However, there was a problem. The lights in my entire home would flicker whenever a character spoke. This was especially noticable when we had a lit up christmas tree that was constantly flickering. It turned out the GPUs we use to power the local text-to-speech were causing "transient spikes" and causing issues with our electricity.
Called an electrician and got it sorted out. With a dedicated circuit installed, the flickering was alleviated, and now we have much more headroom for stacking additional compute.
Onward.