You have to develop subspace (i.e. faster-than-light) communications for this to work. The distance between two points in Europe, after bouncing off a satellite in geostationary orbit, is very, very far, and EM waves can only travel at lightspeed, resulting in very long latency. It's significantly faster if your satellite is in low-earth orbit, but you can't keep a satellite in a fixed position there, so now you need a whole swarm of them so you have sufficient coverage at any time, and you need a way of periodically boosting their orbits or replacing them as they fall into the atmosphere.

There's a very good reason the world likes submarine cables for internet communications.

No, even with Starlink lowering the latency to an acceptable level, there's no way to get anywhere near the total bandwidth of these cables with a satellite constellation, there's limits to how much you can transmit through the air (the biggest issue being there's limits to how tightly you can form beams from your satellite to a station on the ground, and those are quite large, so you ultimately can only get a certain bandwidth per square kilometer).

(This, BTW, is why Starlink isn't and will never be a true competitor to traditional ISPs in all but the most sparsely populated areas: they simply cannot support enough users to make a dent in most places)

No, we won't. It would be easier and cheaper to replace them with microwave links or something like that. Satellites are always more expensive, limited in power, harder to fix, and unlikely to ever match the capacity of underwater cables.

Redundancy doesn't hurt.