DOCSIS 3.0 supports up to 8 upload channels for a total speed of 216 Mbps (on 6.4 MHz per channel). That's shared for all users on the CMTS, which could be a lot, depending on the setup. For comparison, on EuroDOCSIS 3.0, with 32 channels the total download speed is 1600 Mbps for all users, and on DOCSIS 3.0 it's 1216 Mbps.
DOCSIS 3.1 allows for multiple OFDMA upload channels which can allow for up to 2 Gbps shared upload speed capacity, and DOCSIS 4.0 allows for symmetric 10/10 Gbps shared capacity with an undefined split DS:US (you can have the same number of channels). It uses extended RF spectrum in the cable and also needs new amplifiers and splitters.
Which would imply that they should be able to offer a 100/100 connection for the same price as a 200/6 connection and have 6Mbps to spare.
https://www.multicominc.com/training/technical-resources/doc...
(Other sources quote it as 200Mbps? But either is a whole lot more than 6.)
You could sell 1 user 100 Mbps upload with 4 channels, and have one user on a CMTS. You can also sell 6 Mbps upload to 36 users, guaranteeing them 100% of the capacity, or to 72 users, guaranteeing them 50% of the capacity.
It is not feasible to sell packages with more than 50 Mbps upload on most coaxial networks (even HFC, if there are many users, i.e. more than one apartment building connected to a node) as users will not be able to use the full capacity, and the speed will inevitably drop to the sub-10 Mbps level, but the user will be unsatisfied as they are paying for and expecting higher speeds.
To have the same level of oversubscription for upload as download, you would give them all 35Mbps upload with the 200Mbps download. But you don't even have to do that, because who says they all need the same plan? Sell a quarter of them 100Mbps up connections and the others still get 17Mbps up.
No way to justify why it should only be 6 up if it can be 200 down. (And if they're offering 200Mbps down without that much oversubscription by using DOCSIS 3.1 or 4.0, even less excuse for 6Mbps up.)
https://www.excentis.com/blog/differences-between-us-docsis-...
Therefore, given that there is 24x more bandwidth in the downstream direction compared to upstream, it would make sense to offer 20:1 or higher ratios of DS:US speed to customers. Of course, if your HFC network is "good", i.e. has less users, you can sell higher speeds, or just grossly over-sell the service and "hope" not everyone uses the upload at the same time.
Also, a correction: any given cable modem can support up to 32 bonded downstream channels for ~1200 Mbps per CPE, while the cable itself has capacity for ~130 channels, which is theoretically ~4900 Mbps of usable capacity, if there are no TV channels on the cable. However, there is still about ~200 Mbps of upload bandwidth on the cable for all customers, which is limited by the number of channels in the return path RF range. You can assign multiple customers to a single channel, but you can also segment the channels so that all customers get some channels out of the available range, but not exactly the same ones, which would increase the overall capacity of the network.
And yet 200/6 is >33:1, much less 1000/6 at 167:1.
Moreover, to even get to 24x you're using the full theoretical allocation for downstream but not for upstream, which would be 40MHz from 5-85MHz (extended frequency allocation) according to your link. Which would then be ~11x, implying (on average) 200/18 and 1000/90.
Also, who decided it should be impossible to allocate more channels to upload? I realize that's more a question of "who designed this crap" than "why aren't they doing this now" but when the parties responsible are still Comcast et al, that makes it hard to want to give them a pass for it. (Though credit where credit is due, DOCSIS 4.0 supports symmetrical uploads.)