I try to at least give the game a go by RAW before I start making house rules. I figure I should give it a try as written to the best of my ability so I have an inkling of how it's supposed to work.
Often the problem with a rule doesn't show up until you've tried to apply it in several situations.
But sometimes the problem with a rule is completely obvious. For me one of the obvious ones I could bring up was RAW CoC character creation. The core rules don't allow you to create every investigator a player could imagine, and the supplemental rules spread out across dozens of supplements and editions have the problem you'd expect - no unified design and unbalanced options with some professions just getting extra benefits compared to the core one. Plus, EDU is just a god stat in CoC as written that you are massively punished for any roll that doesn't have high EDU. So it should be obvious from the first read of the core rules that something is wrong and become increasingly obvious the more you look at the game that the proposed fixes to the problem are poorly thought out. You don't have to play to realize that.
The same thing is true of just about all of WEG D6. The core rules are obviously incomplete, but the extended material suffers from a lack of editorial control and no product designer issuing hard guidelines about balance. The same is true about 3.5e D&D, where hardbacks were rushed out the door based on desired publication schedule which left no time for quality control. Neither of those games require you to play them to see there is a mess even if the full extent of the mess might not be obvious.
Many times a rules problem is subtle and you won't think about it for years, but truly bad rules stink like a dirty diaper the first time you try to apply them. There are many times one session worth of experience is enough for me to realize it just doesn't work as intended. GURPS, VtM, Exalted, Star Trek 2D20, and Mouseguard all fell apart for me after one session of each. Core mechanics and core play ideas were just problematic in application. The only reason I didn't house rule them is if the core mechanic is bad, it's not worth trying to save it. Goblonia on the other hand didn't fall apart for me after one session, but it still had clear and obvious problems in its rules that I'd have to fix before I'd run games in it.
I have lots of house rules for games I admire. It's not a testimony to them being bad games. It's a testimony to how much they get right and how so much of their problems have to do with the own success, the limits of play testing, the limits of page count, and the need to run a profitable business.
Sometimes I run into a situation where I don't know what the fix is. One that comes to mind is that Web in D&D (particularly in 3.0e which I play) doesn't really work intuitively. It's actually a wall spell, which is fine, but a player is not wrong to intuit that it ought to have a casting mode where you just web a single target. But I've never gotten around to writing the spell to handle that mode.
