Can enumeration take advantage of non-determinism?

If I want to build an NDTM to enumerate a list (of all Turing machines, for example) is there a way to use non-determinism to “speed this up” or take advantage of it somehow?

What types of of r.e. sets are amenable to this–Where having multiple computational branches helps?

Enumeration is a pretty… sequential process.