The following patch is useful for reading multiRefs
within the Body of an Envelope.
I figured out that the multiRefs and the objects which
(may) access them are all siblings within the Body.
Thus, we run into problems when an object/result is
deserialized BEFORE the multiRef.
The patch only re-orders the children elements during
deserialization, pushing up the multiRefs (if they
Since I don't have too much time to dig deep into the
code, I just applied the patch function in the
'decode_value' calls. I hope I did something if not
correct at least valid.