What if, instead, the calibration process worked constantly? Give the user a button to press when they don't hear something well, and another when they do. Let them have this for a month, let them try it in their kitchen, their bedroom, the local shop or bar, outdoors. Run some reinforcement learning algorithm to optimize for getting more "good" presses and fewer "bad" ones. Optionally, adapt separately to each environment.
Is there a "smart" hearing aid calibrator that works like this? If not, but you think it's plausible, I'm interested in working on it.