unichr(2**32) results in a unicode string containing a 0 byte:

{{{
~/mwlib.hg/tests/ python                                            
Python 2.5.2 (r252:60911, Apr 21 2008, 11:17:30) 
[GCC 4.2.3 (Ubuntu 4.2.3-2ubuntu7)] on linux2
Type "help", "copyright", "credits" or "license" for more information.
>>> unichr(2**32)
u'\x00'
>>> unichr(2**32+1)
u'\x01'
>>> unichr(2**32+2)
u'\x02'
}}}

2.6 shows the same behaviour.

Confirmed. 
This happens on architectures where sizeof(long) > sizeof(int):
builtin_unichr() converts its argument to a long, but calls
PyUnicode_FromOrdinal() which accepts an int.

Committed r65339.
Will not backport to 2.5: code that used to (approximately) work would
break.

Thanks for the report!