That's why I dislike difficulty settings that are implemented purely as a fixed multiplier, it can affect things in silly ways.
I think Eroicu had that problem, actually - on normal against the first basic mob (bunnygirls?) you had to learn the punch-punch-block pattern.
But on easy, they died in two punches, so you could zero them out before they ever got a swipe in, fundamentally changing the encounter.
I figure to design a good difficulty switch, you need to take into account what skill your game scales on, which for platformers is often twitch reactions/timings.
By that measure you can scale player health - experts will dodge more shots than novices - but that doesn't help people on a platforming section with insta-death pits.
But I don't believe in scaling enemy health, unless it's a boss that's going to take many rounds to take down anyway.
Instead, my favorite is scaling game speed - I played Eroicu and Kurovadis in fullscreen, because that slowed down the game to about half on my shitty rig.
That way, I had a bit more time to make a dodge or a jump, which made all the difference in making it through the later levels.
You could also get more creative: Add more checkpoints on easy, infinite lives, more platforms, slower enemy attacks, stuff that rewards slow play, ...
The other way around, it can get frustrating when the game punishes you for screwing up: If taking damage cancels powerups, that makes things harder, which causes more mistakes.
At the end of the day, games should be accessible by many people, even those that suck at extreme mario.
There's a lot of satisfaction to be had in games from that "click" moment where you finally get it and suddenly everything works,
but even then, you're going to have people better or worse at the game.
Of course, doing this balance properly takes some time, so the question is how much time people should divert from making the game to figuring out balance mechanics...
