As just mentioned in #75, there is a problem with the spoint3 change from #74:

In version 1.0, the opclass already contained the function g_spoint3_fetch that #74 is adding to 1.3.2.
The function was declared "too experimental and preliminary" in bdc37d1 which is a mega-commit that labelled "pgsphere_1.1.4.916", so later installs don't contain it, but older installations upgraded since then will already have it.

Do we care? The fix would likely be wrapping the "alter operator family" call in upgrade_scripts/pg_sphere--1.3.1--1.3.2.sql.in into a DO block that would ignore the error when the function is already present.