I am using the `same-node` parameter of the “Trigger parameterized builds” so that with jobA→jobB, jobB runs on the same node as jobA. However, if jobB fails and I use Naginator to retry it, it sometimes builds on another node, failing the job.