For example, in Skullgirls you can actually tune GGPO on a sliding scale to add more input delay buffering. You can essentially simulate delay-based netcode within a rollback based game by pumping up the input delay to the max. Players will find their personal preference on the spectrum of "near-zero input lag, more rollbacks" to "high input lag, no rollbacks."
Also, if your game is running at 60FPS you're going to have a lower bound of roughly 3 frames input lag anyway, even if you were playing local. So if you tune your rollback to a 3 frame delay as a default value, the experience for most players is going to be pretty much perfect.
If a frame is 1/60 of a second, that's not much of a design constraint.
in 2000 i had a CRT that could do 75fps, my current screens are 144 and 240 respectively.. 60 is a little bit passé
You want high refresh rate with 60fps tick rate though to avoid dropping frames due to refresh differences.
I mean, I'm sure at the same time you end up with things like the meme Shiki Nanaya win screen[1] but a bad connection is a bad connection will ultimately be a bad connection no matter how you slice it.
[1] https://i.kym-cdn.com/photos/images/original/002/287/289/67e...
This is the kicker. The novelty of having any online play is enough to satisfy most hardcore fans of a game that would otherwise be dead. Melee fans may rant and rave about Slippy, and rightfully so, but they still don't enjoy finding out their inputs were eaten because they got shined ten frames ago. And the fact is, rollback netcode really is dramatically better than delay-based netcode for making precise muscle-memory inputs, and for most fighting games the precision required for inputs is one of the largest contributors to the skill ceiling, so hardcore fans will be happy compared to the alternative. But let's not overlook the fundamental tradeoffs: with delay-based netcode, you will always be able to react to everything that you are intended to be able to react to; with rollback, you will not always be able to react to everything that you are intended to be able to react to. Even if it's a good tradeoff, it's an inferior experience to playing locally... which is fine, because the convenience of playing online is already a tradeoff that plenty of people are willing to make. But if you design your game from the ground up to improve the experience of online play by smoothing out the reaction curve, that will probably come at the expense of offline play in some capacity. Devs don't want to balance a game twice.
You hide it in an artificial buffer of input lag.
Even the GameCube on a CRT has a few of frames of lag. So a high refresh rate monitor can actually beat that, so you can use the difference to hide network latency and still have GameCube-on-a-CRT level latency.