I've filed a JIRA for it here: https://issues.apache.org/jira/browse/MESOS-8169
The change I'm proposing would change validation (for slave reregistration) to only enforce ExecutorID uniqueness in the context of (FrameworkID, SlaveID) as opposed to simply SlaveID as currently implemented. This change might cause trouble for any frameworks that expect ExecutorID to be unique across every non-terminal executor for every framework running on a given slave. That said, ExecutorID is a framework-generated ID and so it's uniqueness *should* be scoped to a framework to begin with. Please comment on the linked JIRA if you're concerned about the proposed change. Thanks, James