This will serve a number of purposes: 1. pre-warm the parser cache and ensure that Parsoid read views don't cause user-visible latency increases, 2. Allow us to do visual diff testing between parsoid read views and legacy read views (T310510), 3. Allow us to do "metadata comparison testing" to verify that parsoid is generating the same lists of categories, language links, and all the other ParserOutput metadata in the post-edit pipeline that core is.
There will be a 'useParsoid' flag in ParserOptions (T300191: Bikeshed name of 'useParsoid' flag in ParserOptions) to fork the cache for this content.