OK, kicked this around a bit and solicited some ideas. How about this?
One of the few things which gets "better" with increasing temperature is VMC - if you are scheduling it with temp, it goes down as temp increases (due to the reduced thrust). So, maybe it's a VMC effect?
Now, I'd usually think that the direct effect of the thrust loss would win; all the VMC scheduling does is reduce the impact of the temperature, but not actually improve things overall. But suppose the aircraft were heavily stop-limited. Then you'd want as low a V1 as you can get, and maybe that drives the balance such that a higher temp, with a (much) lower VMCG, wins out over the reduced T/W?
In-air, if you had an unreasonably high VMCA which was driving adversely high speeds, maybe a similar effect would exist? Though this case implies that the aircraft would be better off simply derating, if more thrust is bad, so it sounds a bit theoretical.
Anyway, I offer up VMC as possibly part of the explanation....