It's likely that there are many ways to turn it off, and without selection pressure they all randomize. If there's 10 codons needed to make it work, and there'

It's likely that there are many ways to turn it off, and without selection pressure they all randomize. If there's 10 codons needed to make it work, and there's no selection pressure, then there's only a one in a million chance it's working by chance. Even when it gets close, there's a 9:1 chance that the next mutation will make things worse instead of fixing the issue. Mostly it would stay around the noise floor with ~2.5 of the codons set correctly. That being said, I doubt it's as simple as "these 10 codons must have these values for this taste receptor to work".