When using a multi-node configuration, workspace cleanup plugin doesn't clean up the left over bits of the job on the managing node.

We have a compile job that compiles on three different architectures. workspace cleanup plugin works fine on each of those three nodes, but doesn't clean up the job that launched them. We are running jenkins from tomcat, using master/slave mode, with # of executors on the jenkins master equal to 0. This causes all jobs to be launched from other nodes. The "launching/managing job" starts other multi-label (in our case multi-architecture) jobs, but it itself is never cleaned.