Gene

In order to agree on time (within the bounds that relativity allows, and for moving detectors that may be relevant to consider as well) it must be obtained from a known source which is comparing itself with other known sources and publishing ex post facto tables of corrections to the time which it was providing.

There is no source for TAI, it is not available until next month, and BIPM itself will recommend not using it.

UTC is available, but its definition is subject to an international committee who are not in consensus about whether it should change its characteristics. Whether or not UTC has leap seconds in the future, it has already had them, so they must be addressed. Also, UTC from one source is not the same as UTC from another source, thus the BIPM publishes Circular T every month, a table of differences which must be applied.

GPS time is under the control of US DoD, USNO, and (even though it is unlikely to change its characteristics at the microsecond level) as such is not palatable for international standards committees. At the microsecond level it is effectively the same as TAI, but at the nanosecond level there are, as with all time scales, lookup tables which must be applied to get agreement. Beyond that, the quality of the individual GPS receivers varies considerably at the sub-microsecond level, and nanosecond precision might require documenting the hardware and software in use during a measurement.

This is an engineering decision which has to conform itself to the available (and subject to change) standards and services. Good luck.

Gene

I believe there are various pitches on how to deal with leap seconds too, although astronomers and aerospace folks may be a better source if to-the-second time-interval determinations are significant.