The main difficulty with split screen is that you need to be able to fit everything you need to render the scene into RAM, twice. Let’s go through some cases:

Just rendering to a higher resolution still lets you get away with the same amount of RAM if you use low-res textures, or a moderate increase because you’re using high-res textures, but only in the foreground – all you need is enough GPU compute power to push the pixels.

If you’re rendering VR both camera perspectives are going to be nearly identical, looking at the same objects, so RAM use is nearly identical to a single camera. Your frame time targets are much stricter in VR, you have to have high and very regular fps or people are going to puke, but again that’s compute pressure, not memory pressure.

In the split-screen case all bets are off: When players are at opposite sides of the map there may be literally zero meshes and textures in common between those two areas and you need twice the RAM for twice the amount of camera views. Nothing in common is the worst case, yes, but it’s bound to happen, and not leave PR in a situation where they have to say “We degraded performance when players are far apart to promote an atmosphere of closeness and cooperation”.