IMO, but it is definitely only an opinion, the higher values of INP means DLM is seeking *more* protection from noise. That was certainly the case before G.INP, and I think it remains the case.
However, the depth of interleaving is only one of the resulting framing parameters that can be changed - the width (or block size) is important (parameter I in the framing data), as are the FEC parameters (R and N in the framing data).
These aren't graphed, so you can't see the history, but my experience is that higher FEC protection was put in place with higher INP values, even if the interleaving depth dropped. I'm not sure of the consequence for the block size. There is definitely interplay between all these parameters.