Ref T9872. Drydock doesn't clean up working copies automatically (so I end up with lots of copies of the same thing), and the UI doesn't give me any decent way of seeing which working copies are unlikely to be used for a while and can be manually released.
My build hosts run out of disk space because of this, and it's a nightmare to fix (because I have to SSH through an SSH tunnel through a VPN to get to the build hosts to clean them up).
As per T9872, many of my repositories have submodules, which dramatically increases the size of a non-bare working copy (caching the bare repositories is often much smaller, which is why I wrote the original working copy blueprint like that).
Since these build hosts reside on physical hardware, I can't just magically add more disk space like on EC2.