Agreed. Players want their 20-Str barbarian to never lose in a strength-based contest against a 5-Str wizard because they don't like being shown up by a wimpy bookworm, but on the other hand they want their 20-Str barbarian to have a chance to avoid being grappled and swallowed by a 35-Str purple worm. You can't have it both ways; flat distribution or bell curve, full stats or stat modifiers, an X-point difference is an X-point difference either way and you can't make the same type of roll benefit the underdog sometimes and penalize the underdog other times. Decoupling stats from rolls, using different types of rolls, etc. just shifts the problem somewhere else and doesn't solve the underlying issue.
Depends. Some people think it's a feature when there are things that a 1st-level trained character or a mid-level untrained character can't possibly do that a mid-level trained character can do and a high-level trained character can do trivially; some people think it's a feature when there's still a chance, however slight, for any character to do anything; some people think it's a feature when everyone's baseline for success is fairly close together and different stats/features/etc. can push you off that but not too far. 3e is mostly the former, 5e is mostly the latter, 2e/4e are somewhere in between.Then I suppose that it's also a feature that the chance for him to do so goes up by each successive edition? Because the bonus for being trained has gone steadily down ever since 3E, and 5E also reduces the spread in ability scores.
The first perspective works best from a PC vs. NPC perspective (if you're pitting your weaknesses against someone else's strengths with that big of a gap between them, you probably made a bad decision somewhere and you shouldn't have a 25% chance to turn out just fine) while the third one works best from a party vs. world perspective (if the variance in a certain stat is too large then challenging parties containing an expert PC, a mediocre PC, and a sucky PC with respect to that stat can be hard), and the second one is a compromise between the two, so it's really up to you where you want that tradeoff to be