Do we know at this point whether switching to pandas 3.0 would even help
with the Python 3.15 build failures? I wasn't expecting it to be
needed. I'll try it in Salsa, but that will take some time.
The pandas build failure is some datetime operation returning None
instead of the object it's expected to return, which isn't one of the
potential issues I know about. I don't yet know whether it's an actual
bug, or some other package that just needs to be rebuilt before pandas.
(rpy2 and statsmodels also failed to build, but I'm guessing those just
need to be done after pandas.)
