"Sub 2:50 was misplaced optimism..."
Never said sub-2:50 was misplaced optimism. Also did not sate sub-2:50 was inconsistent with workouts noted.
"You must put in/expect ridiculous training."
The opposite. What was stated was OP did more long runs and at faster paces than when I dropped to 2:50.
Also observed most would log a series of shorter long runs incrementally peaking to 20+ -- and not do six or more 20+ mile long runs. And that most would not approach MP for these long runs.
"Should her longs have been actual marathons..."
Hyperbole. See above. Does anyone suggest marathon races should serve as training runs? Or are required as fitness indicators?
Also noted...
* 2:50 represents a significant improvement over best marathon result, 3:27 (age graded this is a jump from about 64% to 79%. )
* to be reliable indicators, training runs should have reasonably certain times and distances, such as by being supported by an uninterrupted GPS track.
* races, particularly longer lead up races, are good indicators - but no consistent race results were provided (until now).
* it is risky to pace for sub-2:50 without solid indicators, hence an initial suggestion to conservatively split "1:28 or so".
* a buffer for negative splitting is always a good idea, but more so when fitness is less certain
* a 5K is not a good indicator of marathon fitness, but a sub-17 result supports a sub 2:50 when combined with training run details... also that a sub-17 5K time is faster than some women running under 2:50 are capable of.
* based on disclosure of 5k result, lowered suggested halfway split from "1:28 or so" to "1:24 - 1:25"