But GPS satellites have constant velocity so calculating this is pretty easy. For a future super probe that accelerates and decelerates and maybe passes through many different gravitational fields (giving a general relativity time dilation) and perhaps does so for many centuries, the possibility of error mounts up.

I'd vote for time being defined in terms of some universally observable phenomenon. Say the rate of cooling of the CMB. Anyone with a super spiffy radio telescope anywhere in the universe could then tell what time it is.