I initially liked the idea but...

First, to the average non technical person (including many amateur photographers, unfortunately) ICC profiles are unknown.
For the more technical inclined ICC profiles would seem more like EQ-curves as used in pre- and deemphasis. And that's not what MQA is about. MQA's data reduction might be compared to video subsampling but for the time smearing fingerprint we need another analogy, I'm afraid.