In the post-processing step (logstash?) identify where the error comes from at a high level (such as extension name or ResourceLoader module name) and ping graphite so it can show error frequency graphs for those error sources.
This has been resolved for Minerva/MobileFrontend (T202026).
We could probably generalise that by moving the simple mw.track -> statsv counter to WikimediaEvents instead so that we have it for other skins/desktop as well, possibly fragmented by skin id.
But, I don't know if this is something we want to further invest in, given work has now started on using EventGate and Logstash for full error indexing (T217142) which would make this obsolete.
At the time of creating this task it was assumed that the server side implementation of error logging would be based on Sentry. We have eventually decided on a different implementation, so de-tagging.