Improve support for client apps to monitor SDK health
"Monitor the monitor"
On Edge, we've run across a number of issues where the SDK was in an unhealthy state (telemetry not sending until a restart on long running sessions, silently failing on opening database/ses). Ironically, one of the parts of code we have the least telemetry for is this SDK, which is critical for uploading the other telemetry.
The SDK does expose some information through the "debug events," but I think there's an opportunity for a better system (ideally, something that works well with aggregate metrics). Alternatively, maybe the "stats event" could be reworked, but for Edge it would be preferable to be able to use our current pipeline (rather than setting aside separate storage/processing just for the stats event)
I'd consider this and #225 the most impactful issues for Edge, and they're both fairly large chunks of work.