Based on some of the Manara et al papers, especially the one regarding the Christiansen wavelength, I think an IR (wrong) COP of 9 could be done. The pore size and grain size of alumina would have to be optimized to concentrate the peak IR emission band significantly more than is usual for typical alumina materials. If someone wanted to fund this, it may have industrial applications, like heating specific materials without heating the surrounding environment much, sort of like the way a microwave oven is optimized for heating water molecules, but using IR radiation instead.
Possibly using a band pass filter on the IR camera could accomplish something similar.
It would be hard to do a false COP of 9 by accident using IR, though, in my opinion, without also measuring electrical input poorly due to wavelength manipulation or delta-wye mix-ups.
Yes, that reminds me:
For input power mis-measurement, using a current clamp above the current it saturates will do that... So that is another option to add to the known reversed clamp or the trivial measure one phase and forget to multiply by 3. Delta-Wye switching does not directly cause an error with a 3 phase power analyser, but could contribute indirectly, or if power was measured in some less robust way, for example by measuring only voltage.