To measure surface temp, yes.
The remaining question is what is the reasonable amount of effort to go to, to get a still not quite applicable reading.
I'm wondering if you even meant epoxy, or did you mean grease?
again... of the surface temp.
You are arguing towards an ideal that is non-applicable to the final goal. It will be a higher temp, it maybe a certain % more accurate, but the remaining issue was whether the additional effort was worthwhile since we are not trying to determine an epoxy breakdown number.
If you want a viable heat reading you must at least measure the temp at a short length from a good thermal junction. The top of a plastic epoxy isn't that.
Just, NO. If there is a runaway situation and you're measuring epoxy temp, your device is probably dead already by the time the runaway trend has occured long enough to be seen as a runaway. Setting a peak shutdown threshold would work better, and having it automated would be necessary. It'll still need rely on a more accurate measurement method than surface temp of an epoxy carrier- remember the context there, these are chips on a hard drive PCB, not something dissimilar.