There are plenty of things which are excellent ideas, but don't make it into the kernel or major distreaux... or once there, fail to endure.
FatELF is one particular example, which failed because of a combination of technical and political issues in the (GNU-administered) toolchain and the Linux kernel... I forget most of the detail by now.
I note that x32 does have a Wp page, but I think the really crucial thing is the posting (and subsequent discussion) at
https://lkml.org/lkml/2018/12/10/1145 . Via a comment from the maintainer (JPAG, who ISTR is another HPC buff) I see
https://popcon.debian.org/stat/sub-x32.png which suggests that x32 currently has fewer than 1,000 users... scaling that by the overall popularity of FPC in the wider Linux context I'd speculate that only one pascal user is interested :-)
As I've said though: it's an interesting question. I wonder whether there are any other architectures which similarly provide a split word/address size ** , and whether you would be able to make a good case to the FPC core developers for its being to the advantage of the project if they were supported?
** Leaving aside IBM "big iron", where all discussions of support got bogged down in the merits of EBCDIC over ASCII, or in modern development tools' deficiencies when asked to do something simple like providing a cross-reference listing.
MarkMLl