Why do C- and Fortran-ordered NumPy arrays show the same L1 cache-miss percentage when formatting a row?

Part of

Runtime Debugging Challenges In Developer Toolchains