So, bad class design? If bad class design is the root cause then any fix that does not start with an adjustment to the class or the base rules that the class relies upon is inherently a wrong design change.
The problem lies not with the classes alone, but with the interaction between those classes and the ability raises. I'd argue that the fundamental problem is
not the class design, it's the diverging stat bonuses. To put it another way: that any fix that does not start with an adjustment to the ability modifier divergence or to the base rules that the ability modifiers rely upon is inherently a wrong design change

.
In any case, builds that "share" stats start off at a disadvantage; and that's a viable choice - it's OK. What's not OK is that these effects become so overwhelming as the game progresses.
On the topic of one-trick pony's...
I hope it's not an intentional design aspect, but I also don't agree that there are only two shades here. I'd also not have sympathy for a character that tries to be good at everything. That's probably even worse than a one-trick pony because it's unachievable. Can we just try to get a character who's good at one or two things (though not super awesome best-in-the-world good at one thing) and just doesn't suck at everything else? You know, average? Instead, the "one trick pony" is awesome at one thing and sucks at everything else.
Personally, I don't see the attraction in playing one trick pony's in the first place. But some of my players do. They'll come up with amusing builds with odd tricks, and bore of them quickly, and then switch character (though, this hasn't been much of a problem in 4e yet, to be honest).
These PC's are generally not problematic in actual game-play. They're usually quite fun and spicy precisely because they're not ordinary and not bland. I think maybe we're not talking about the same thing here - I'm not talking about rule-bending cheese (maxed out to-hit with storm of blades, orbizards that make saving virtually impossible). I'm talking about focused characters that know what they're good at.
I thought this was answered in my post. IMO, it "is the beginning concept of a one-trick pony." It may not HAVE to be, but it usually is.
An 18 is the default stat distribution of any auto-built character builder character. It's hardly a one-trick pony, rather, it ensures that at the very least, the character will not be terrible at hitting. It does not much impact your FRW defenses. Rather, a character with an 18 tends to be
less one-sided than others: If you're into char-op, raising a few other stats for specific feats' prereqs may be wise, but a basic 18 character won't have access to a few specialized feats (e.g. polearm gamble in heroic tier). Really, as long as it's not disrupting gameplay, what business does a DM have micromanaging the PC's choices like that? And really, a stat distribution is like the least disrupting game element I can imagine in 4e. We're not talking about a barbarian with iron armbands and a bloodclaw weapon abusing storm of blades here...