Here are some preliminary angle details.
1.5deg is quite small and Oth orders out to thousands and probably many thousands of beams which will require some pretty mega processing on the satellite. Even so, 1.5° is still quite a bit of area on the ground. At 540km 1.5° directly overhead is over ~150 km^2 and out at the edge of coverage its easily 2x that.
Just doing very round math (or using C-student intuition), its clear that the strength of the constellation is going to be the massive number of satellites with numerous overlapping coverage areas, so beams from multiple sats can simultaneously cover the same areas. This is really where Starlink is going to excel in total user base over a constellation like OneWeb or Telesat, where there will be significantly fewer layers of coverage over any particular place.
Round Math Assumptions:
--One sat is 20gbps
--One beam gets all of that (reality is there's no way this happens, but it at least gives a good bookend)
--Average coverage area of 200km^2 (compared to the best case 150km^2)
Good Math Willing, the somewhat generous assumptions above basically serve each km^2 with an average of 50Mbps or, for imperial visualization, ~0.2Mbps per acre. So, if you lived in a rural area where every user had a 5 acre plot, each of you would get an average of 1Mbps download from one sat. That doesn't sound great, but that's actual average traffic (not like a speedtest rate or anything), so its probably not the terrible-est.
The good news is that even at the current ~1000 satellites users in higher elevations have upwards of 3-4 satellites worth of coverage, which you can kind of hand wave as 3-4x that 1mbps rate. For the ~4000 satellite constellation the layers of coverage won't scale linearly, but its probably a safe 8-10x for much of CONUS—at least the northern half--for a round number average of ~10mbps/user for the 5 acre example.
I'm not a network expert here and welcome a cross check, but that actually sounds pretty reasonable when you factor that network traffic typically has has a lot of peaks and valleys--average rate doesn't tell the whole story. To analogize, your main house breaker rating isn’t simply the sum of all the circuit breakers, its much much lower than the sum.
Variables that will affect the above numbers:
--Most importantly, the beam hopping scheme. Its really hard to imagine the entire sat capacity being put into one beam, both from a hardware perspective and also latency—especially if there's thousands of beams to hop around even a 1ms per hop leaves a user waiting a long time for the next burst of packets... Anyway as previously noted I am NOT an antenna person or a network person, but I'd surmise that there needs to be many tens if not hundreds of active beams in order to keep the latency manageable. This would be negative linear scaling of the above numbers; if there are 100 beams, the above math more or less linearly factors down from 1.0mbps/user to 0.1mbps/user.
--User density of course makes a huge difference too: if we're talking a user per 1 acre (which really isn't rural anyway) the above math scales linearly down to 2mbps/user, if we're talking a user per 20 acres the above math scales linearly up to 40mbps/user.
--Additional satellites: This won't exactly be a positive linear scaling, but from a practical perspective it is pretty unbounded--Starlink will almost certainly be able to scale up as much as they need to support the plausible global user base Sats at higher altitudes will have less performance than the 540km case played out above since each beam will cover more users, and of course the opposite is true for sats at lower altitudes.
--Higher Performance Satellites: Like any technology product, SpaceX isn't going to just keep sending up the same satellites. Based on user demand and technology evolution, they're going to get better, with some combination of narrower beams and more simultaneous beams to better isolate users, basically serving a smaller number of users with the same bandwidth.
--Higher Performance UTs: It is likely system performance will be bounded by PFD limits, which basically means that a satellite cannot just blast users at a higher power in order to have a stronger link (and thus higher data rate). While a UT doesn't necessarily need to form a ton of beams and also doesn't need mega processing power for complex hopping (it pretty much just needs enough beams to track a handful of satellites across the sky one at a time) higher directivity and gain will also generally improve link performance. I wouldn't see this as a major upgrade path for the system as the UTs would have been sized to begin with for pretty great performance, but certainly there will be evolutionary upgrades that may not manifest as outright faster speed but will certainly improve the overall service.
--More channels: V band is the next great thing for satellite comms; this will be a step function when it makes it to users. There’s plenty of FCC and ITU filings out there in the >Ka region...most are go-nowhere real estate grabs, so it’s not like they’re going to be gone forever.