The criticism of MAPIE is incorrect, there is no reason why larger house should be more difficult to predict just based on the fact that it is larger.

Prediction Intervals in general should reflect complexity of prediction of individual objects and it could be that small house is much more difficult to predict price for based on specific dataset.

On the other side there are some methods that produce constant width of conformal prediction regions (making them the same for small and large houses) if that is the case MAPIE should certainly consider adaptive methods such as Conformalized Quantile Regression.