After gh-23838 we now default to building with Meson (via meson-python, when invoking a build with pip install . or similar) in main. There are a number of more niche build configs not yet ported, and a few things were xfail-ed and need fixing up. Here is a list, roughly in order of priority:

• Build nightlies for Python 3.12 to unblock other projects from testing with 3.12 (xref CI: BLD: build wheels and fix test suite for Python 3.12 #23991)
• Improve how CBLAS and ILP64 flags/dependencies are handled (xref BLD: undefined symbol: cblas_sgemm when building with meson with pip install #23909, and also automate passing CFLAGS/CXXFLAGS - I'll work on this next): see BLD: improve handling of CBLAS, add -Duse-ilp64 build option #23984
• Finish implementing all BLAS/LAPACK auto-selection and control options (e.g. the equivalent of NPY_BLAS_ORDER). Done in ENH: meson: implement BLAS/LAPACK auto-detection and many CI jobs #24893
• Fix up the issue with PyPy and re-enable the pypy3.9 CI job on Azure. Done in BLD/CI: re-enable ILP64 usage and PyPy job in Azure #24238.
• Integrate SIMD support (the relevant CI jobs are still going through runtests.py)
• Update the build docs (partially done in DOC: Update building docs to use Meson #24573; the rest in DOC: add "building from source" docs #25350).
• Fix and re-eanble the 32-bit Windows wheels builds with MSVC on Azure (or move to GHA, or drop? - see DEP: stop building and testing 32-bit windows #23717). Re-introducing on GHA in CI/BLD: fail by default if no BLAS/LAPACK, add 32-bit Python on Windows CI job #24279.
• Fix the long double representation check so the builds works on some of the more niche platforms again (xref BUG: New meson system fails to identify long double representation on 64-bit BE platforms #23972)
• Fix the xfail-ed test_cython for Windows due to a missing npyrandom.lib (done in BLD: change file extension for installed static libraries back to .lib #23983)
• Move Travis CI jobs over to Meson
• Check that all CI jobs have been moved over (e.g., by removing runtests.py and setup.py, then running CI and ensuring nothing fails). Done - there were a lot left. See Restructuring GitHub Actions CI jobs #24410 for the tracking issue for CI migration.
• Investigate how to fix the bin/f2py3X CLI install, in addition to the f2py and f2py3 ones (see project.scriptsinpyproject.toml`, and commit f22a33b). Or can/should that interface be dropped? Multiple installs into a single env is considered bad practice by now. Dropped in BLD: only install the f2py command, not f2py3 or f2py3.X #24235.
• Investigate and fix the two einsum test failures with int8/uint8 that were xfailed in BLD: switch to meson-python as the default build backend #23838 (EDIT: these two tests fail with clang-cl on Windows too, see ENH: Add changes that allow NumPy to compile with clang-cl #20866 (comment)). EDIT: opened separate issue: BUG: einsum with int8/uint8 input is buggy #24732
• Fix the array-api entrypoint installation (see xfail in numpy/tests/test_public_api.py). Done in TST: enable test that checks for numpy.array_api entry point #24249.
• Fix the longdouble string repr test failures on aarch64 - see Issue printing longdouble on Linux aarch64 with the Meson build #23974 and Revert "TST: disable longdouble string/print tests on Linux aarch64" #23985
• Move the Pyodide CI job over to Meson
• Clean up the apparently unused skip statements under [tool.cibuildwheel] in pyproject.toml; the platforms to build for are controlled in .github/workflows/wheels.yml
• Compare all files included in sdist and wheel between the setup.py and Meson builds, ensure nothing went missing
• Fix numpy.__version__ (see comment further down)
• Remove usages of numpy.distutils in user-facing docs. Here is one example.