In re player facing mechanics vs Complexity
I'm pretty sure this is a case of correlation vs causation. AFAIK there's nothing inherently requiring these to be lighter on crunch, but in practice they generally are. Ditto for FKR.
Unisystem Lite (Buffy, Angel, Army of Darkness, Ghosts of Albion) is slightly crunchier than D&D 5E core with all player facing rolls, but this requires a good bit more prep for the GM, as the core mechanic is essentially always opposed rolls, but NPCs (including monsters) always roll a 6, and this is bound into their action value.
Dragonlance 5th age is less complex than D&D, but more complex than its mechanical sibling, Marvel Super Heroes Adventure Game; DL5A has 9 suits, 8 atts, and is player facing, while MSHAG has 5 suits and 4 atts, and is mostly opposed play of cards. This makes MSHAG much easier to use the PC gen sequence for NPCs, too...
I can't get past how buffed the characters are in 5e. I like the action economy very much, and I think skill checks are good, but overall, the game as written produces superheroes by level 5, and that's just not my jam. The last campaign I ran, we used some rules variants (gritty realism, low availability of durable magic items, slow level advancement, took out some spells) and it was better, but...still no. I like DCC and BRP!
I've always run D&D as Fantasy Medieval Super Heroes. That's always how it's come across to me. I didn't realize that for the first decade, but looking back, I ran it as FMSH.
Re WFRP
Gotcha. I've never played Warhammer, but I read a bunch of the Gotrek and Felix books and loved them! What is the core mechanic in the game?
Which edition?

WFRP 1E: 1d100 ≤ (either Attribute or 10× attribute, by attribute). M, S, T range 1-10, W 1-20, and A about 1-5. WS, BS, I, Int CL, WP, FEL 12-120. Skills fall into categories: +1 attribute, Eliminates a penalty, provides a +10 or +20 (by which skill), provides a special ability. Careers allow a maximum alteration from original rolled attribute while in them.
WFRP 2e: all atts nominally 1-100; skills avoid a halving at first rank, and add up to 20 with additional levels. Careers similar to 1E. Fewer atts. Shifts damage from d6's, d4's, and d3's to d10's and d4's.
WFRP 3E 1-15 range for atts, fewer still, different career system, uses custom dice (a few too many types)... good, but also tied to component heavy playstyle and collectorism as much of the material in older core was only in supplements. And the supplements all added more cards. I enjoy it, but it's got issues.
WFRP 4: basically, like WFRP 2e, but with skills to a max +40, and instead of +5 per attribute gain and (US = ×½) +0/+10/+20 for skills, to 1 per raise on each. Combat almost eliminates missing. Careers far more linear than any prior.
Soulbound: d6 count successes on 6. Not the same setting as the others.
To answer the OP's question
For me, the «compile time» vs «runtime complexity» is a significant element.
«Compile Time» complexity
More and more, I dislike mathy «compile time». It's the numbers of calculations more than the complexity of the individual calculations; GURPS is over my tolerance and has been since about 1995.
The limit for me is Rolemaster: the edition I have, using only the ChL/CaL, SL, and AL/CL, with RMC 1 ONLY, is fine, to a point. No secondary skills. It's more math than my players currently would tolerate. It's just a lot easier to use a classless system. In play, it's very much classes don't matter - they just affect the cost of skills and spells. Hefty «compile», fast and simple «runtime» (provided 3 digit +2 digit chain addition isn't a problem - a simple 4-Fucntion calculator per player is a great play aid) and one doesn't use tableless combat (RMC 1), Breaking 150, nor the separate initiative/action costs options.
Hero System I'd run again if the idea of multiplication and division of fractions didn't scare half my current players. If I were to use it, first use would be a mundane Semi-Mythic Greece. It has more complex calculations than GURPS, but about 1/3 to 1/4 as many for non-(supers/casters/monsters).
I really like the character gen systems for: Star Trek Adventures, Firefly, all the YZE games I've read except Forbidden Lands (which is probably why I've not tried to get FL to table), Sentinel Comics, Street Fighter, and VTM 1e
I like but find cumbersome the CGen for Prime Directive 1E, Dresden Files, CORPS 1E/2E, MegaTraveller.
«Runtime Complexity» my sweet spot is Mouse Guard - there is exceptionally good advice on difficulties (if a bit rigid), the conflict mechanic is super flexible (if a bit too abstract for some), most uses of skills are mechanically covered in simple ways. Plus its GMing adventure format is a brilliant bit.
Year Zero engine is fine in every flavor I've run; 2d20 likewise.
Prime Directive 1E is the limit of math crunch in play for me. My current players would find it a bit too mathy. (doing 3 column single + {single or double} digit additions/subtractions. Frequently)
Rolemaster or Spacemaster in play are table heavy; that's an issue, but one I'm good with. It's a solvable one, but it's still an issue. MERP was better - fewer tables, but the same general lack of actual in-play complexity aside from tables.
If I need a calculator in play, from time to time, I don't mind. If I can't run the game without one, too much.
D&D 5E is, for me, way to many special cases on an otherwise solid system. The player base, the tropes common, and the directions the publisher has been going have moved it from "alright, I can run it if y'all want" to "Don't ask for it. I'm done with them."
I've always liked the medium-light rules, but don't mind higher complexity when it pays off in play.