> which effectively halve the available bandwidth
I disagree (and I'm not sure where this idea is coming from-- do you think they use the same spectrum or something?) I think they effectively "add" bandwidth for the system by allowing satellites that do not have saturated uplinks to "share" capacity. E.g. a satellite which is mostly over the ocean but can see a ground station can now be more useful.
Yes, sat-to-sat enables new use cases. But it also has other effects.
All the satellites above me have an uplink from a ground station.
Also, satellites above ocean with few people in their footprint can see my uplink but can't use it effectively, because they don't have many users in their footprint (possibly none, because of elevation angle restrictions).
If we have satellite-to-satellite links, then those satellites above ocean can share their bandwidth with satellites above me. Therefore, real system capacity goes up.
Assuming that ground stations are upgraded as the size of the constellation grows-- they're strongly directional. Their throughput is the uplink bandwidth * the spectral efficiency * the number of satellites within LOS of an uplink ground station. Intersatellite links let you use more of this throughput without it going unused and "spoiling"-- that's the main point.
> which may not be the case if the satellites around (for sat-to-sat links)
No real sat-to-sat links yet. The exact specifications aren't disclosed, but they'll likely be 100gbps symmetric links.
> the base station that it sees are maxed out.
You can expect spectral efficiency for the uplinks to be better because of big antennas, and for them to be provisioned with no problem illuminating all the satellites overhead.