Good point raised by @adamgundry in some other project:
[..] I would guess that Ridley is reporting a stale metric value that is not being updated because of the exception, does that sound plausible @adinapoli? if so, is there some way that if a metric update fails we can stop reporting the metric altogether, rather than reporting a stale value?
This has to be investigated, as I cannot remember how (and when) ridley exposes the metrics for Prometheus to pick 'em up.