babel/plural.py contains the following line:

    ('ellipsis', re.compile(r'\.{2,3}|\u2026', re.UNICODE))  # U+2026: ELLIPSIS

But \uXXXX escape sequences in regexps were added only in Python 3.3.
In earlier versions, \u stands for literal \u.
In other words, in Python 2.X this line is equivalent to:

    ('ellipsis', re.compile(r'\.{2,3}|u2026', re.UNICODE))

...which is certainly not what you wanted.

NB, the re.UNICODE flag affects only semantics of \w, \W, \b, \B, \d, \D, \s and \S, so it is no-op here.

The dubious regexp was found using pydiatra.