There is yet one memory management bug (with unconsumed_tail).

flush() does not update unconsumed_tail and unused_data.

>>> import zlib
>>> x = zlib.compress(b'abcdefghijklmnopqrstuvwxyz') + b'0123456789'
>>> dco = zlib.decompressobj()
>>> dco.decompress(x, 1)
b'a'
>>> dco.flush()
b'bcdefghijklmnopqrstuvwxyz'
>>> dco.unconsumed_tail
b'NIMK\xcf\xc8\xcc\xca\xce\xc9\xcd\xcb/(,*.)-+\xaf\xa8\xac\x02\x00\x90\x86\x0b 0123456789'
>>> dco.unused_data
b''

What should unconsumed_tail be equal after EOF? b'' or unused_data?

>>> import zlib
>>> x = zlib.compress(b'abcdefghijklmnopqrstuvwxyz') + b'0123456789'
>>> dco = zlib.decompressobj()
>>> dco.decompress(x)
b'abcdefghijklmnopqrstuvwxyz'
>>> dco.flush()
b''
>>> dco.unconsumed_tail
b''
>>> dco.unused_data
b'0123456789'
>>> dco = zlib.decompressobj()
>>> dco.decompress(x, 1000)
b'abcdefghijklmnopqrstuvwxyz'
>>> dco.flush()
b''
>>> dco.unconsumed_tail
b'0123456789'
>>> dco.unused_data
b'0123456789'