TL;DR: After implementing array argument broadcasting, I'm not aware of
any deficiencies in the updated SRFI 231 that I think need to be
remedied.  I still need to document the changes, but otherwise I like
the library design.

With this commit:

https://github.com/gambiteer/srfi-231/commit/36c7522136a06f6b58b664366084b0cc49f18186

I implemented implicit array argument broadcasting for the procedures
array-{map|for-each|every|any|fold-left|fold-right}.  Following Python's
NumPy, but not Racket's math/array, array arguments to array-stack and
array-append are not implicitly broadcast.

Whether array arguments are implicitly broadcast is determined by the
parameter array-broadcasting?.

We do not do what Racket's math/array calls "permissive" array
broadcasting, largely because the array transform requires a non-affine
map that doesn't compose well with other transforms.

There is not yet any documentation about the changes, but there are
minimal tests.

Brad