I am back sooner.
David Hess, that would be any user's interpretation, and indeed should be so. I agree with you entirely.
Wolfram, yes sir it appears to be a strong assertion, but I do have long experience to qualify. In doing it, I admit to being guilty of overlooking heat dispersion through the core, which you have wisely brought my attention to. Since modelling for this is well nigh impossible, I have just conducted a test to see how I can correct myself for the mental lapse.
This Staco 125VA 2A 120/132V variac has a winding tap at the top end, to offer a small step-up. I simply set up a a constant current of 1.50A dc in this 'overwinding' for 90 minutes. The main part of the winding was left open altogether. (I chose not to make any connection to the wiper brush, because of uncertainty of measuring resistance through it). Measuring the hot and 'cold' resistance showed a temperature rise of just over 68C over a local ambient of 27C. The unexcited part of the winding was barely warm. I did not stretch this to a heat run at 2A for fear of magic smoke.
The extrapolated temperature rise for 2.0A would be 120C over 40. Much too high.
The suggested aluminium plate heat sink was not attached during my test, and should give some relief, my guesstimate is 100C over 40. This may be barely acceptable if the wire, and all adjacent insulation, can be classified as F.
It was never my purpose to apply higher loads while operating at mid voltages; on the contrary, it would have pleased me to be able to use the full rating in amperes over the whole range of settings. I now know what happens!
'Generic' variacs may not be very different: however low one is hung, the one has to use the same copper wire; and class F insulation is not hard to come by. I will do a simulated heat run on a couple more variacs.
Gyro, you are right too, the extra heat at the brush contact is highly localized and can cause a hot spot. Fortunately, the brush is gripped in a massive block of brass to heat sink it. Hopefully, heat which gets on to the winding track will spread out well enough through the copper. Please suggest a good practical way of testing what margin if any exists here.
Thank you all once again.