This infrastructure exposes a bunch of RTS related counters/gauges (allocated mem, number of GCs, etc), as well as a bunch of metrics related to the various endpoints of the backend (using servant-ekg). I took the opportunity to also serve the standard EKG UI to easily be able to peek at the evolution of things of interest, but we also can access all those things from a JSON API, which we'll be able to use later to grab metrics before/after some operations, to have a way to observe and quantify performance improvements -- the UI uses that same JSON API.

Here's what the UI looks like. (Note that we can dynamically add graphs for any of the metrics tracked by the system.)