For line scroll, you have to check scroll values once per scanline, but for column scroll, you have to check 320/8 times during one h-blank interval, while the scanline beam is racing across the screen. It's very resource intensive, hence why it is limited to 8px. But it could be 16px not 8px, I don't remember.
Hard to tell why Darius Gaiden slows, it could be anything from the CPU or the VDP1 not being able to keep up, or just a stylistic choice for that matter.
Saturn VDP2 has:
- a single BG colour (overscan),
- a single colour per every scanline
- two rotating backgrounds (scale/rotate)
- four normal backgrounds, but only two of them can scale or do line/column scroll.
Enabling the second rotating BG will disable all four normal backgrounds, so it's not useful. I only remember two games using it.
For the first rotating background only, you can set two rotation parameters. So you can have the top of the screen tilt at a different angle than the bottom, but the two parts never intersect. For example you can do an infinite ground plane and a non-tilted background, which rotate in unison and join at the same "horizon".
So the system only really has like 2.5 scaling backgrounds, so to speak.
ST-V is a stock Saturn, but running cartridges. Imagine it as if the Saturn was running an entire game off the cart slot. Keep in mind that the VDPs can only draw what has been copied into their own memory pool, so you can only display so much frames per second before you hit fillrate caps.
Rotating the framebuffer in Rayforce would only rotate the sprites, not the backgrounds. The game makes extensive use of line scrolling backgrounds, and these are the ones you lose when you have to rotate the screen.