I grew up primarily playing games with a soundblaster in OPL mode. I got a wavetable card fairly quickly, but even at age 10 or 11 my reaction was uhhhh... I can never find the picture but you know the one that's like "when squaresoft does HD remasters" and it's a cat sprite with "high definition" fur textures that looks like shit? yeah it was like that.
most game music was intended to play through total shit tone generators and when you add 'better' samples and effects it just sounds embarrassing. like, "midi-driven player piano in cheap sushi bar" type embarrassing, where the high quality of the sounds is rendered comical by the total lack of expression of the score being played through them.
In recent years I've wavered a bit on this. It used to be that if I wanted to listen to the doom soundtrack and i pulled it up on youtube, if it was played through a Roland anything I'd roll my eyes and go look for an OPL version. That's changed; while many of those still sound absurd, some actually appeal to me more than the FM versions I was used to for so long. still, a lot of it just completely misses me.
Ironically I never owned a Gravis Ultrasound, but I've listened to GUS-intended versions of game soundtracks on youtube, and Roland-intended versions, etc. and... man, I have to admit, I have the same response that I do when I listen to games that came out on the FDS that used the extra sound channels - namely, I think they mostly sound terrible.
I believe those songs were composed specifically for the extreme limitations of the original FC hardware, and then the FDS channels were added, not the other way around - and what they put in those extra channels just guts the impact of the music for me. It just sounds awful. I don't like FDS Metroid or Zelda or Castlevania II (CV3 is good tbh) because they all add elements that just dilute the melody, take the punch out of it.
This is also often how I feel about "enhanced" DOS soundtracks, and for the same reason: anyone composing for the PC would have been an absolute fool to target the Roland, a thing that like 50 people had. People had adlibs and things with adlib compatibility, and I would be astonished if hardly anything wasn't composed for those first, then "upgraded" for Roland and GUS after the fact. I am not opposed to artists taking second stabs at things (god knows I do it enough), but I do believe that, very frequently, ones initial "final draft," the first version of a work that they'll let anyone see, has the most heart and soul that they can manage; a later attempt to stick the landing better isn't doomed to be less sincere, by any means, but I think it's definitely a big risk.