Currently en.wikipedia.beta.wmflabs.org runs master of VisualEditor and a manually-synced version of Parsoid that is the same as production. This is great for testing VE as to whether it's good to deploy into production, and had served us reasonably well so far.
We can't just run Parsoid as master, because (a) Parsoid master is occasionally not compatible with VisualEditor master for days or even weeks at a time as the APIs change, and (b) unless Parsoid is going to be pushed live at the same time as VisualEditor (and split into a het-deploy situation for the cluster's wmfX vs. wmfX+1 deployment) we wouldn't know if VE and Parsoid were truly OK to go out.
However, bug 56583 showed that we also need to test master (and ideally also current production) of VisualEditor against *master* of Parsoid, in Labs, for automated browser testing and so on before any code goes live.
Do we need en.wikipedia.beta.wmflabs.org to have en-Parsoid-master-VE-master.wikipedia.beta.wmflabs.org and en-Parsoid-master-VE-prod.wikipedia.beta.wmflabs.org duplicates? (Eww.)
Also, are there other teams who need their multiple independent components and cross-dependencies? Wikidata, perhaps?
Version: unspecified
Severity: normal