To determin if we are correctly and effectively providing context for selser, we should collect statistics for the different, mutually exclusive cases layed out in the parent task (T318393):
# Page creation. Selser does nto apply since there is no previous content.
# The page exists, but no base revision ID provided. Selser is not possible.
# No render ID (etag) provided by the client, falling back to the latest rendering for context.
# Render ID (etag) provided by the client, found in stash.
# Render ID (etag) provided by the client, found in parser cache.
# Render ID (etag) provided by the client, but not found (edit fails with a 412).
Keys:
```
htmlinputtransformhelper.no-original.*
htmlinputtransformhelper.no-original.page-exists
htmlinputtransformhelper.no-original.page-not-exists
htmlinputtransformhelper.get-original.*
htmlinputtransformhelper.get-original.in-body
htmlinputtransformhelper.get-original.with-renderid.*
htmlinputtransformhelper.get-original.with-renderid.not-found
htmlinputtransformhelper.get-original.with-renderid.found
htmlinputtransformhelper.get-original.with-revid (may be a cache hit or we parse on demand, we can't tell)
```
Later, we want to break up the with-renderid.not-found and with-renderid.found keys:
```
htmlinputtransformhelper.get-original.with-renderid.*
htmlinputtransformhelper.get-original.with-renderid.*.*.not-found
htmlinputtransformhelper.get-original.with-renderid.*.*.found
htmlinputtransformhelper.get-original.with-renderid.stash-hit.found
htmlinputtransformhelper.get-original.with-renderid.stash-miss.paser-cache-miss.not-found
htmlinputtransformhelper.get-original.with-renderid.stash-miss.paser-cache-hit.found
htmlinputtransformhelper.get-original.with-renderid.stash-miss.paser-cache-hit.not-found (render key mismatch - works for counters, but it's a terrible name!)
```
NOTE: This should be covered by a test case that asserts that for each request, exactly one of the above keys gets set.