The following code was added by commit 964e02a:

+                                       /* Skip float and double, could be
+                                          "unknown" float format */
+                                       if (ptr->format == 'd' || ptr->format == 'f')
+                                               break;

When this change was made, 'd' format was implemented with _PyFloat_Unpack8() in little endian and big endian tables, and _PyFloat_Unpack8() had a if (double_format == unknown_format) code path.

The if (double_format == unknown_format) code path was recently removed by commit dae85c4 in the main branch.

So yeah, it makes sense to copy the little endian or big endian function as the native function in the _struct module.