Sadly, I found a bug within the 6LoWPAN/6Lo code af the release when trying to
integrate it with RIOT :-( (find fix attached):
Since 3a8af61, when setting the destination address for a 6Lo device the
network interface's hardware address is set instead of the previously fetched
address from the neighbor cache.
Also mind that my fix technically isn't correct for IEEE 802.15.4, as mixed
address formats (e.g. source 16-bit, destination EUI-64) are possible, but
since nd6_get_next_hop_addr_or_queue() does not return any information about
the hardware address length, there is no way of getting the actual length of
the destination address (as far as I know). This is why I decided to just
revert the regression introduced 3a8af61, but to add a comment so this might
be fixed in the future.
Sorry that I didn't provide this fix during already for the release candidate.
I noticed something was wrong when testing it, but as always, time is short
and found it only now to properly pin-point the issue :-(.