Introduce a systematic health-tracking model for adapters

[geeknoid]

We need to handle adapter failures in a systematic way. Whether an adapter panics, always returns errors, or constantly times out, we need to have a clear detection, tracking, and recovery plan. If we try to "reboot" the adapter a few times and things still fail, then we can take the adapter out of the running permanently until the operator tells us otherwise.

We need to architect this health-checking & tracking environment for maximum robustness at runtime. Let's spec out the model and implement the support needed.