Some set of conditions is causing p4-plugin to occasionally think it has to sync an entire tree on slave machines, instead of incrementally updating from the last sync. Based on the behavior of p4 on these slaves, it would appear that p4-plugin is wiping the #have list for some reason.

The build in question is a Build Flow + Job DSL setup with parallel jobs, some of which take advantage of setting the pinLabel parameter for CL consistency with previous steps. This runs on a homogenous build farm. Shelving/unshelving is also used via external (command line) scripts. Also, the root build uses one workspace, while all the downstream builds in question share the same customWorkspace, and use the StaticWorkspaceImpl + SyncOnlyImpl.

This is especially painful for me as my p4 tree is 1TB, which takes a very long time to sync.