This just happens to be one of my hot buttons.
The problem is the assumption of NIST traceability in the calibration of dynamic torque. The problem is that calibrator response time, and torque driver speed/momentum are not the same between torque drivers/calibrators (different brands, and even different units (same mfr/model)). This is not accounted for in specs of the calibrators or drivers (electric/pneumatic).
So yes, NIST traceability is not necessary for a calibration. In a former job, I spent many years in the early days of the semiconductor industry. In that context, NIST traceability was non-existent for many parameters for many years. In such instances, a "de facto" standard, "golden unit", process sample, or other means was used to provide consistency in the process.
We are in agreement about this philosophical detail.
In the context of dynamic torque (as someone with some background in statistical process control), my concern is that there is an absence of "something" to provide consistency. In an electric torque driver, for example, variations in line voltage may change the effective rotational speed, and thus, the actual force applied to the screw/nut/etc. The same may hold true for a pneumatic driver. On the calibrator side, the response curve varies between transducers (think of it analogously to the risetime/bandwidth of an RF device). Two transducers which read identically at a given static torque will not necessarily equate with each other at a given dynamic torque (I've seen this experimentally and practically).
So although it is fully acceptable to have defacto/consensus standards where traceability does not exist, I am only propagating words of caution regarding the pitfalls in the area of dynamic torque.
The reason I am bothered about calibrations to dynamic torque is not the lack of traceability, but the lack of consistency or repeatability.
The most knowledgeable in this area (as I understand) refrain from calibration of dynamic torque because they are aware of these pitfalls/inconsistencies, and instead rely upon static torque to set critical parameters.
Again, I would sincerely love to hear from a torque expert on this, as I have found it quite bothersome.