These are precisely the discussions that lead to the fundamental question, “For whom Fortran, for what!?”
Sure if I am looking out for myself for my practice of Fortran with only my limited view of a “frog in a well”, looking at my own “toy” programs, or as a one-person team developing solutions - even if mid-size - for use by one or a few other people, and with a certain number of objects coming into existence during run-time such that their sizes are individually manageable, sure my own tedium with having to author such “allocation” subroutines can be ignored, especially by others who strongly prefer to view Fortran narrowly for computing needs that require FORTRAN 77 plus implicit none and may be free-form source and C interoperability and perhaps coarrays.
But as I mentioned previously (point 1), there are computing needs where the calculations and simulations involve multiphysics architectures that are spread across many teams with many contributors working with innumerable number of objects in memory as part of countless classes (like derived types in Fortran) whose allocation sizes can be parameterized using a few quantities, similar to nobs and nvars above. This is precisely where the value of the length-type parameter comes in and where much tedium can be avoided.