I actually think that the matrix in Shadowrun is amazing.

It just doesn't add to the larger game. A negative mechanic can be tremendously clever, but it just detracts from the game. A positive mechanic would be something like D&D 5e's advantage mechanic: it's a really cool way to give flavor and is super-simple to execute, so if you aren't opposed to it for some reason it's a really nice addition to a system that is pretty simple.

The problem with the matrix is that it doesn't just adapt a trope into the game, it does so in a way that makes a whole separate game within a game.

Regarding the notion of the matrix versus astral projection there's a little bit of a false correlation there; the astrally projecting magician definitely sees different things, but they're still fundamentally in meat-space. The matrix often has a connection that way (especially in installations where the matrix is embodied by a physical representation of that space), but data-runs, for instance, are going to almost always take the matrix people to an entirely isolated sphere, while a projecting mage is at least in an overlay of reality.

It's probably fair to say that some of this is a problem with Shadowrun, but it's a problem that comes from a design philosophy that is willing to accept a certain amount of fragmentation and then jumps in wholesale.

I will agree that it's possible for some groups to have a good time with Shadowrun's matrix mechanic, but it's also been something that's definitely proven to be negative more often than positive; as a game designer you want to look at the average, not the exception, and the problem is that Shadowrun doesn't have a system for the average roleplayer; it has a particularly math-heavy game.

For early Shadowrun, where there was much less user-friendliness in many ways (at least, less of an attempt at such) than in 5e, this was fine (I also think that there was some rules elegance that got lost over the years, but that's for a later date). Each character had their own nuanced elements, though deckers were almost always the most difficult in my opinion (barring really weird exotic builds).

The real issue is that you have a game that has some interesting design decisions, but which has become too clunky and bulky. A decker doesn't enjoy a separate ruleset, they use the full ruleset and then a superset (or, often as many as three supersets, since they will have cyberware and sometimes rigging), and requires the GM to do so as well. Compare this to a magician, who might also have a couple supersets (e.g. spellcasting, summoning, and projection), but will often share those duties with other characters because, unlike a decker, the other supersets don't represent a way to recover sunk costs but rather another path to excellence.

As a designer, what I'm always looking at is when there's a part of a game that reflects a problem within the philosophy or design structure. That doesn't mean that any part of the game is necessarily bad in its own right, but it could be something onerous. To use another FASA-affiliated game, the Battletech roleplaying game has life-path creation systems (and only life-path creation systems) in the Time of War reboot that my players just couldn't sort out. It's a fantastic system once you get to play, but if you don't have a lot of buy-in it just becomes a problem that players don't want to deal with.