The failed Windows job is not related to this PR. It looks like some sort of CI infrastructure issue:

[...]
 Downloading markdown_it_py-3.0.0-py3-none-any.whl (87 kB)
   ---------------------------------------- 87.5/87.5 kB 2.5 MB/s eta 0:00:00
Downloading pygments-2.18.0-py3-none-any.whl (1.2 MB)
   ---------------------------------------- 1.2/1.2 MB 37.4 MB/s eta 0:00:00
Downloading urllib3-2.2.1-py3-none-any.whl (121 kB)
   ---------------------------------------- 121.1/121.1 kB 3.6 MB/s eta 0:00:00
Downloading zipp-3.19.0-py3-none-any.whl (8.3 kB)
Downloading mdurl-0.1.2-py3-none-any.whl (10.0 kB)
ERROR: THESE PACKAGES DO NOT MATCH THE HASHES FROM THE REQUIREMENTS FILE. If you have updated the package versions, please update the hashes. Otherwise, examine the package contents carefully; someone may have tampered with them.
    unknown package:
        Expected sha256 7a89180babf3b6af321f5d6970f8a90d687977565a48f05dc4ce75dfb94c777d
             Got        52c33cdda6efdc34dc995f43c7b0cd170e99b63331e9f2bc1d7a2aad6471df30

Error: Process completed with exit code 1.

@rgommers, do you have an example of how this could cause a problem? Clearing the WRITEABLE bit dynamically has nothing to do with const-ness of arguments in compiled code.

Clearing the WRITEABLE bit dynamically has nothing to do with const-ness of arguments in compiled code.

You'd think that yes, based on how things should work - but tell that to Cython. The example in gh-17172 does what you say here, and it was failing for a very basic linalg.expm call. Cython chokes on read-only arrays unless arguments in the Cython function are declared as const.

Heh, I even commented in gh-17172 (a mind like a steel sieve). OK, I'm not going to pursue this further.

If odeint was a new function, we could specify up front that the arrays passed to user functions will be read-only, so if a user implements their functions in Cython, they must use const appropriately. But I guess we can't add this now without introducing a backwards compatibility break. Maybe it could be considered for a SciPy 2.0 API change.

Maybe it could be considered for a SciPy 2.0 API change.

Agreed, and even then it's a stretch perhaps. At a minimum we first need to fix all our own Cython code to work with read-only arrays, and properly test that.