> something with AR-15 DNA that is more modular

Modularity isn't something "AR-15 special". SOPMOD and, well, videogames merely popularized it. Something like AK isn't inherently unmodular - e.g. AKV-521

> What exactly is deeply flawed about the M4?

Design flaws common among AR-15/18 and most of their derivatives:

- tight, unstreamlined cyclogram, with notably short BCG overtravel. Even minor deviation in conditions, ammo quality and state of the weapon can cause some kind of failure.

- no primary extraction. Major common cause of failure. It was perfectly understood back in the 19th century, yet AR designers still failed to implement it.

- pre-engagement. Common сause of failure in adverse conditions. Among AR derivatives, only AUG has a crutch fix for that. Garand fixed "RSC M1917", Stoner "unfixed" it.

- primitive bolt-action style BCG/magazine interface. It does not take into account mag dynamics and necessity to lift rounds into the bolt path in split-second. Major cause of failure to feed exacerbated by bad magazines and very short overtravel mentioned earlier.

- plunger ejector. No competent and self-respecting gun designer would ever put this crap into a military rifle.

- inherently weak bolt design that routinely fails in two different ways. Seven locking lugs make zero sense mechanically and structurally, a design indefensible from good military gun engineering practices standpoint. Every tiny lug is a weak point, and most bolts (milspec included) have square lugs with stress points in every corner. There is also a big hole drilled through the bolt - and it is stressed, with bolt cam pin as a separate piece. This is why there is a whole genre of photography called "pics of AR bolts broken into pieces". *Some companies make custom AR bolts with rounded lugs that at least exclude corner stress points. Australians made a 3-lug bolt which is a way more sound design.

- bolt cam pin that is offset pretty far back. This causes unwanted torque in the bolt. Contributes to reliability issues.

- loose and wobbly Schmeisser magazine well, which was a major cause of issues even in MP38 already, exacerbated by design dimensions meant for disposable aluminum magazines. It does not secure the magazine snugly, as it inherently requires clearances, and those clearances still can be gunked up preventing mag insertion. Same with damaged well or damaged mag. Inserting on closed bolt requires force, and there are infamous over-insertion and wrong insertion issues. Disposable mag debacle caused perpetual mag durability issues that remain unresolved. In general, AR magazines are the major pain felt by every military user.

Now, some more M4-specific:

- Aluminum receiver breaks, literally.

- buffer that has to be precisely tuned for the rifle and ammo, and has to be in good condition? In a military firearm? Seriously?

- linear main spring arrangement may be nice for a competition rifle, acceptable for a marksman rifle, very questionable in a general purpose military rifle.

- multiple non-captive small parts. Lack of proper gunk accumulation and friction management. E.g. parts like firing pin are unfluted. Lots of obstruction potential. Common cause of failure.

- design reliance on brass cased ammo, which is heavier, much more expensive, and prone to shortages in a war, as brass is in limited supply.

- I won't say anything bad about DI gas system, but this one does fail in water, catastrophically.

- M4 has no proper alleviation of multiple inherent short barrel issues. It is loud, flashy and fast-firing.

These are off the top of my head, not all of them. These are the important ones.

Overall, "1,691 rounds between failure" is atrocious, and M4 demonstrated even worse than that in adverse conditions. It fails to work completely in some of them, fails catastrophically in some others.