Sometimes slaves share resources that hudson is unaware of. For example, we have two linux build servers that host several chroots with variations (amd64, i386, different versions of linux). For each chroot, we have a hudson slave. It would be nice if those slaves could be grouped together, and allow a maximum of X executors per group (instead of per slave). That would avoid the common case of it running multiple jobs on the same physical machine, but in different chroots, while leaving other physical machines idle. Maybe make this a "soft" limit - avoid running in the same group if possible, but if not possible, allow it.

(Note that this same scenario could be applied to multiple physical machines each hosting multiple VMs; in fact, we have a physical machine hosting a windows vm and a linux vm, and then multiple chroots inside the linux vm).