Relevant configuration - I have one cloud setup, with 4 different AMIs defined. They are each technically identical, the only difference being that they have different Jenkins labels so that certain builds only happen on a certain machine. Let's call them A, B, C, and D.
Expected behavior - If all build hosts are stopped and I start a build for project B, only the build host for project B will start.
Actual behavior - If all the build hosts are stopped, and I start a build for project B, the A build host is started first, and once that is up, the B build host is started, and once that is up the B project starts building on it as expected. This introduces a longer than expected delay when starting builds.