You cannot do any kind of bandwidth shaping, priorization or fair
queueing on any link but the bottleneck.
that is plain bullshit.
I think you are talking about two different things here:
Henning Brauer is explaining how queuing and prioritization work in OpenBSD,
while the others have in mind those network switches and routers which perform
packet switching in hardware. In such devices, input frame or packet is
forwarded without any queuing if the output port is not congested. Otherwise,
the queues would quickly grow and overrun (and the latency would be higher).
This is the only possible thing to do when there are tens or hundreds of
interfaces, especially high speed ones.
So, it is probably not a good idea to automatically apply knowledge about
network hardware to packet switching in general-purpose operating systems.