This is a rather complicated answer. I hope after you read whatever I end up writing below that you are not more conufsed that before.
From a QS9000 perspective, most definitely any M&TE that are used to test product. Additionally, the MSA requirements surrounding those. You can draw the line there (maybe). If it affects product quality, etc., then it is in that classification.
The second echelon (I'll call the gray area), are those M&TE that are used in maintenance. The choices are (In my view - I am a metrology professoinal, not a QS9000 auditor)
a. calibrate all of them (easy to pass an audit that way, but not necessarily most cost effective).
b. Define those M&TE (still in the gray area) that maintain equipment used in processes, or to maintain other parameters that COULD affect something. This would include multimeters, handheld test equipment, environmental monitoring test equipment, etc. Those also all have to be calibrated.
The third echelon are those M&TE that are most definitely outside the scope of the process. These do not have to be calibrated. But I am stickler enough, that if any company wants to go that route, I strongly advise having it well documented. In some cases, it is almost easier to calibrate all of it than to prove that it cannot affect the process. Some examples are the scales used in shipping and receiving areas, multimeter used by a technician who strictly works on refrigerators in the cafeteria, Thermometer used by the cook in the cafeteria, test equipment used by people who service company vehicles, a nice spectrum analyzer used by the telecommunications tech who works on the LAN equipment (in a semiconductor mfg environment, this would obviously not impact the product - or scope of registration).
My philosophy has been to calibrate anything that measures. And if a decision is subsequently made NOT to calibrate a quantitative piece of test equipment, I still have it in the database. I even maintain a file in the calibration records. I put a memo in the file detailing the who/what/where as to why it is not calibrated.
If I had a meter lab, and 50 meters on-site that could affect the product, and also that could affect the effectiveness of the preventive maintenance done on anything in the factory that could conceivably affect cost, quality or on-time delivery of the product, they would all be calibrated. If in that situation I had one other similar model multimeter that did not affect anything, I would opt to calibrate it.
If it were a case of a significant cost to calibrate, I would weigh the business decision. Then if it did not impact anything that mattered (as above), I would opt to not calibrate.
Any decision by a QS9000 or ISO9000 (moreso in QS9000) company to pot not to calibrate, in my view should be thought through carefully.
Okay, guess I have muddied the waters enough for now. I'm sure there may be those who disagree. But hopefully I have provided some useful insight that may be helpful to someone.