I just moved the npm jobs in the MediaWiki queue over to the new,
faster, docker based CI system. Instead of requiring building a fresh VM
for each 20-30 second job, it uses a docker container. I'm expecting
this to make a noticeable improvement in CI response times.
For the most part everything should just work, if something doesn't,
please file a bug in the #ci-config project and CC me (legoktm).
If something has gone terribly wrong and I'm not around, you can ask a
CI team member in #wikimedia-releng to revert  to put everything back
to the way it was.