As part of the build, we do run a bunch of tests and these tests do stuff on the image and I suspect they leave stuff that interferes with subsequent runs.

I want the ability to configure agents to only run one build.