it's obviously not a "speed" issue as you point out. In the past there were a reasonable number of "slow" 5,000/10,000 guys who ran under 2:10. Ron Hill's best was 28:30 something, Clayton's 28:40 something, same for Dick Beardsley. Ian Thompson's 5,000/10,000 times were 14:05/30:10 when he ran under 2:10 though I believe he later got into the mid 29:00s. Neither Henrik Jorgensen nor Gerard Nijboer were ever under 28:00 for the 10,000. The Japanese crank out loads of guys who go under 2:10 without getting near 28:00 for the 10,000.
There is some correlation, obviously, between fast marathoning and fastish 10,000s. You're going to have a hard time running 2:10 if your best 10,000 is 31:00. But once you have enough 5/10 speed to go under 2:10, or whatever time you want, the advantages of getting faster at those distances diminish. It doesn't hurt, consider Steve Jones. But it only helps so much.
So what needs to change? All of the old time guys I mentioned had the marathon as their best event, but they were really distance runners who mixed the marathon in with serious racing at other distances. It seems to me that our guys now often wait until they've maxed out at track distances and then move to the marathon rather than mix the marathon into their racing in peak years. Lynn Jennings did that. Mark Nenow did it. Ryan Hall did it and only Hall really ran the marathon as fast as his potential indicated and then faded out. Shorter was serious about track, cross country and marathons all at the same time as was the case for Hill, etc.
The other thing I see with many of our current marathoners is that they don't stay with the event if they aren't successful pretty quickly. Sometimes you have a great performance in your first race or two but sometimes you're like Hill and need several years to go from 2:24 to 2:09. I think we'd get more sub 2:10s if our guys ran 2-3 marathons a year for several years even if they aren't all spectacular ones.