I have two qualifiers before I make any comments.
1. My library is packed, so I am taking information from memory.
2. I am not an MSA guru (I expect to be by the end of the year).
My recommendation is to get a copy of the MSA manual from the QS9000 set, if you do not already have one. Internal to my company (a fortune 100 electronics company), we have a spec written on MSA requirements, which is essentially a paraphrase of the above MSA manual.
If you want an R&R study of torque wrenches, you need a torque standard. I suppose it depends on what type torque wrenches you are referring to. If they are the type that each user adjusts to a needed setting at each use, that is one case (I'll call those "variable torque wrenches). If they are fixed "click-over" type that are set to a specific torque setting by a calibration lab, and used for a calibration cycle at that setting, that is another case (I will call those "fixed torque wrenches").
For variable torque wrenches, repeatability I think would be to set at a given setting, and make the repeated measurements. Reproducibility would be to return them to a very low setting, then let multiple users on multiple days and shifts set them for the desired setting and make the redundant measurements.
It seems that what is needed is a torque calibrator. You would need to have multiple operators set them up as above, don't let the operator see the readings on the torque calibrator, then have operator set for desired setting, and turn until they click over multiple times on the calibrator (making sure that they cannot see the readings, as this may bias their operation of the wrench. To keep the study blinded even further, take the torque calibrator out to the production environment without notice, and ask operators with wrenches already set as they would normally use them in production to click-over on the torque calibrator.
For fixed torque wrenches, same as above, only the wrenches are already set to the desired torquing force.
If you don't have a torque calibrator, I am not sure what other instrument could be used, other than something that retains a fixed amount of torque.
The alternative if you do not have access to a torque calibrator, is to get hold of manufacturers documented specs for the torque wrenches. Those specs should be based on statistically derived tolerances. If it improves your comfort level, you could contact the manufacturer, and ask for some of the data. If you use their data and/or specs, that may be useful to determine cpK, etc. of your processes. I think even if you have the above data/specs, there is the added consideration of operator-to-operator variability of torque wrench setting. In other words:
a. If each operator set torque exactly the same (doesn't happen in real life), how much variability is there in the actual application. That would be repeatability.
b. Added to that is, how much difference is there between the actual setpoints operators set the torque wrench to.
Just a few thoughts. Again, I am not a guru. But I hope I have helped more than confused.
------------------