My view is that if the system breaks when the players play it ("game" it), then it is a bad system.
More generally, if playing the system doesn't produce the fiction, and the experience of the fiction, that is intended/desired, then it is a bad system.
To put some flesh on the preceding bones: in my experience, the RPG which reliably produces the biggest gap between intended/desired experience, and experience of play; and which is most vulnerable to breaking when players actually try and play it; is AD&D 2nd ed. And I think it's possible to be fairly concrete about what the issue is: the mechanical system as presented sets out chances of success (to hit, damage, non-weapon proficiencies; thief skills) and also gives some players (the players of spell casters) some buttons to push (spells in the per-day spell load out); but the intended/desired play typically involves the GM controlling when the players succeed or fail (in order to manage the unfolding of the prepared story) and hence disregarding or setting aside those chances of success and mitigating the pressing of the buttons.
This won't work if the rules are transparent to the players!
The 1982 Traveller Book has the same issue. This is from p 123:
The choreographed novel [my emphasis] involves a setting already thought out by the referee and presented to the players; it may be any of the above settings [ship, location or world], but contains predetermined elements. As such, the referee has already developed characters and setting which bear on the group's activities, and they are guided gently to the proper locations. Properly done, the players never know that the referee has manipulated them to a fore-ordained goal.
This sort of thing won't work if the rules are transparent to the players.