My understanding is that the starlink satellites will use T-Mobile’s spectrum to communicate with the cellphones, which requires a partnership.
This seems like a stupid idea on so many levels.
It's up to the network operator to efficiently manage their spectrum in a way that does not interfere with other radio users, and mutually between their own base stations.
If T-Mobile/Starlink manage to do that, you get additional connectivity in remote areas on your existing devices. If they don't, and their own network quality degrades, you can switch networks; if they degrade other networks' quality (actually, not just as a hypothetical), hopefully the FCC will shut them down.
This is an extremely power limited application. We're talking signals just above the noise floor here.
Think beam-steering MIMO WiFi where a base station can direct the vast majority of its energy directly toward a client, meaning clients in other directions never see the transmission and so the noise floor is lower and bandwidth higher. Similarly such systems can "see" which direction a transmission comes from and so distinguish multiple clients transmitting simultaneously - much the same way your eyes have no trouble seeing two people waving at you from different directions even if they're twins and dressed identically: the spatial information is sufficient to tell them apart.
To a rough approximation the delta between the theoretical limit and practical limit is computationally driven these days and we've gotten really good at the computation part.