I would be interested in that but I have a feeling you would wipe out the memory and CPU. It would be cool of you to try. A method that *would* work would be to divide the bandwidth evenly between all those subscribers by bumping up the number of flows from the default, and using a tc filter by destination IP address in a single htb + fq_codel instance. In that case any subscriber could use up all the bandwidth left available and in the worst case degrade to total fairness. I do not think that is what you want, but a differently sized shaper per customer (?)
In that case preseem, LibreQoS, etc, can accomplish what you want on a cheap 20 core Xeon box instead and take your splynx output to run fq_codel (preseem), or cake (LibreQoS). I only have numbers for LibreQoS but you can easily pump 25Gbits for 10k subscribers through that without breaking a sweat.
https://github.com/LibreQoE/LibreQoS/re ... /v1.4-rc10 runs on Ubuntu, is open source, and the latest release is available here:
https://github.com/LibreQoE/LibreQoS/re ... /v1.4-rc10