oversubscription is a necessity in the economics of last-mile consumer grade ISP services no matter what the tech: ptmp wireless last mile, docsis3 cablemodem on coax, gpon/10Gpon fiber last mile, dsl, active ethernet fiber last mile, geostationary consumer grade ku and ka-band vsat, starlink, etc.

if you look at the traffic usage charts from 50 houses on residential last mile gigabit connections the average customer really doesn't move that much data. you'll get the occasional outlier such as a person who is really enthusiastic about torrenting 75GB 2160p movies but when averaged out over a group of dozens or more customers it's generally a non issue.

the danger is in oversubscribing your network to the point at an unreasonable ratio that the ordinary end user starts to see bad packet loss, jitter, latency, slow speeds, etc.

Starlink announces that they will make service available at a certain quality for $X/month, but more people want that than they can sell it to. Basic economic theory says they should raise the price to find equilibrium, but consumers hate that kind of thing, tying (in some cases) the company's hands.

I don't really think that Starlink's hands are tied with respect to price increases, though. I doubt over-subscription is all that possible, unless they overestimated their own capacity when signing existing contracts.

Too many people, not enough bandwidth.

Latency, timeouts, general slowness. There are a lot of layers that starlink needs to optimize/scale, they're effectively trying to operate as a global cell provider.

Insufficient density or spectrum to serve all users in an area