I followed your steps as listed and actually got it to where it works without the WAN crashing. I think I got too aggressive on the ingress and egress and that was the main cause. I started at 113Mbps and 10Mbps and tweaked the advanced cake configuration with 'nat dual-srchost' and 'nat dual-dsthost' strings in the queuing disciplines for ingress and egress respectively. The advanced cake tweaks recovered quite a bit of fairness between my devices. I got it working with an 'A' for buffer bloat at ~190 Mbps down and ~19 Mbps up. The average speed on dslreports.com test is now 130-145 Mbps down and 19 Mbps up. Would it help further to reduce my ingress to 140 Mbps from ~190 Mbps in regards to performance and the buffer bloat? Thank you for all your help. I am appreciative you were able to give me a solution to this issue.
Thank you for your help. It seems I was too aggressive on the ingress and egress and that seemed to be the problem. Still not sure why I had to reset the modem though, but if it happens again with my new settings I will run the commands you listed and paste the output of logread.
Well, so that seems nice then. But again the crashes certainly are not expected, if sqm runs out ouf CPU cycles the worst thing that should happen is that either it does not control bandwidth or latency under load as well as it should, but certainly it should not crash. So I would try to investigate the crashes/wan losses in more detail as they seem to be orthogonal to the sqm configuration issues.
That is quite easy to test, and I have no a priori hypothesis, so just do it?