Only just finding time to respond to this now, sorry.
Yeah, I realize this is a problem. But I feel like it’s probably more of a surface-level issue born of people adapting existing mechanics exactly, than a terminal flaw for the concept as a whole, and nothing that can’t be sorted out via a combination of prototyping and smart pacing/level design.
In particular, I probably wouldn’t put 8 enemies on the screen at a time. I think close melee combat against multiple enemies is often either unfun, or a justification for Batman/Mordor-style fake combat where you just mash attack and respond to counter QTEs. (Not to say those games are bad – Mordor in particular is about more macroscopic things like crowd management, dealing with captains, navigation, stealth, etc, so it makes sense to abstract away the combat.) Even some shooters, a genre where more enemies makes sense, seem to throw a hundred enemies at you because “it’s a videogame” instead of doing what makes sense wrt the mechanics and the story. (Looking at you, Tomb Raider reboot.)
If I felt like the game needed an army of mooks for the player to plow through beat-em-up style, I’d consider some combination of: not allowing them to block, giving them guard meters, giving the player character attacks that are effective against the guard of the mooks but not serious opponents, giving the player character assist moves (e.g. stun grenades) that allow them to preoccupy other mooks while they deal with one directly, etc. I’m also partial to not letting enemies walk through eachother – whilst it kinda exposes some of the absurdities of 2D combat, I feel like allowing opponents to overlap each other and all get hit by the same attack is also pretty absurd.
If I couldn’t avoid placing enemies on either side of the player character, I’d probably start considering something like a block button. The thing you risk losing here is directional inputs/blocking, which I kinda want. One idea I’m contemplating is a parry/catch-counter button which, if you also input a direction whilst pressing/holding it, makes you face the opposite direction, so you’re encouraged to parry an opponent coming at you from behind.
An advantage to modeling the game off of KI instead of SF is that KI is into the idea of strong forward-moving attacks, which could allow you to put some space between yourself and someone behind you. Giving the mooks tamer, normal-driven movesets a la SF could make their approach comparatively cautious, which buys the player character space. Special moves which strike on both sides and deal some pushback would probably also make sense, and again, assists which allow you to preoccupy a mook on one side while you take on another on the other side could also work.
So I don’t really have a single answer, I think it’s a combination of these things which requires some prototyping to figure out. But one thing’s for sure: I’d want to really play up one-on-one encounters with serious opponents and/or boss figures. That’s kinda where you can get into nitty-gritty frame trap/mixup-driven pressure, footsies, zoning, yomi and all that, which is sort-of what I’d want the player to be building their character towards, and I generally wouldn’t want that getting broken up by mooks. There are probably interesting encounters that can be built involving two or more opponents operating synergistically, too, but that’s kind-of a “see if it works in prototyping” thing as far as I’m concerned.
I’m pretty sure DH made a KI card game to prototype the combo system. Rare might’ve made one back in the day, too.