> The desire to do this is, actually, *why our make supports multiple
> jobs in the first place*.  You see, our make comes from Sprite (or at
> least by way of Sprite, at one point long ago in its history) and on
> Sprite, you just type "make -j foo" and foo jobs are started, then
> silently migrated off to other machines in the cluster as happens to
> best optimize load.
> 
> Gosh, it'd be nice if NetBSD could do that. ;-)

sprite had process migration, and its clusters were tightly coupled.