I used to use the queueing feature of OpenBSD's PF a lot. It worked pretty well, but it made a monstrous mess of my firewall configuration. Not really PF's fault - I had a lot of rules and adding the queue statement to each one made it hard to read.
Since being easy to read is a pretty important quality for a firewall rulebase, I disabled all my queueing stuff. That's a bit of a shame, but I needed to be able to understand how my firewall worked.
I have recently set up an offsite store for my most valuable data, and I'm using Duplicity to securely back it up. That's nice, but since it's a huge upload, that kind of traffic really screws with the performance of my ADSL line. For the uninitiated, filling the upstream bandwidth with data leaves no room for TCP ACK traffic, which stalls pretty much all TCP traffic.
Anyway, I set up a simple priority queue for ACK traffic. I expected an increase in (e.g.) browser responsiveness while doing the upload, and for sure I got that. What I didn't expect was an enormous increase in the upload speed! It pretty much doubled it. I'll buy that for a dollar.
While I may not go back to my old, CBQ-based scheme with all its confusion, the ACK priority queue is a winner.