Rewind a year ago and contributing to Treeherder was much simpler. Install vagrant/virtualbox, run `vagrant up` and you're pretty much ready to go. However now that we have to build the UI for it to work at all (due to the use of JSX; bug 1336556), the workflow is more complex, more error prone, not as well documented and some things just don't work. For example: * to do front-end only development you have to appease yarn's engines check (ie have the correct version of node installed), otherwise it gives an annoying error (bug 1416257) * to do full stack development, you have to juggle both `yarn start` and `./manage.py runserver` and know which port to use to actually view the thing * login doesn't work in many combinations and you have to read to docs to figure out which (bug 1363722, 1317752) * source maps aren't working, so debugging the site in the browser is a pain * `yarn start` in vagrant doesn't auto-reload since it's not using the poll option * building and running `yarn start` take ages * (until recently) it wasn't possible to run the new JS/selenium tests in the vagrant instance because firefox/geckodriver wasn't installed (bug 1340305) * Pulse ingestion doesn't work without having created an account manually on pulse.m.o and adding the credentials to the vagrant environment * ...and so on We really must make Treeherder development painless again.

Oh and also in the last year or two: * more and more jobs are moving from buildbot to taskcluster, so the buildbot only `ingest_pulse` is less useful * Vagrant seems to be increasingly unreliable, whilst Docker is looking more and more polished