This report is for python.el in the current master
(c10828bd8520db83cc06a2ad5de6f8a1ad74b83c).

Pressing C-M-a one time correctly positions the point at the beginning
of the current class or defun statement. Pressing C-M-a multiple times
is futile, as the point gets stucked where the first movement left
it. This is not consistent with the usual emacs behavior and it's not
very convenient to navigate code anyway.

The problem is in python-nav--beginning-of-defun:

            (when (and (> arg 0)
                       (python-info-looking-at-beginning-of-defun))
              (end-of-line 1))

This puts the point at the end of the line when the point is at a defun
or class line and the navigation is backward. So the next
re-search-backward will match the current line. This is fine when the
point is *after* the column where the defun or class statement starts,
but not quite so when the point is *over the indentation* whitespace
(that is, before the beginning of the defun/class keyword).

One simple solution is to add another condition:

            (when (and (> arg 0)
                       (> (current-column) (current-indentation))
                       (python-info-looking-at-beginning-of-defun))
              (end-of-line 1))