I discovered today while playing around with fasttrack-compatible queue trees that they do not work when L3HW offload is enabled on the fasttrack rule. It seems obvious in retrospect. Unfortunately, disabling L3HW fasttrack on the CRS309 leads to high CPU demand even at modest throughput (300-400mbps). Might want to make a note in the OP about that in case people are banging their head against the wall trying to figure out why their queues never get hit.
@mducharme I’m curious to hear what you think of the recent QoS-HW improvements in RouterOS 7.10 and how we could take advantage of hardware marking with fasttrack-able queues in the future.