i.e. if the original numpy array x is C-contiguous (which is the norm in numpy) then x.T is F-contiguous (which is the norm in Julia) so PyArray(x.T) is F-contiguous and can in theory support fast linear indexing for example (i.e. is not strided).
I’m not sure if PyCall.PyArray actually supports fast linear indexing but PythonCall.PyArray does.