Let's take 2:01:09. That's 4:37.4 per mile. Let's say the question is if he could run 3:59.0 for a single mile. That would require his MarathonPace to be 16.1% slower than MilePace (In the sense that 4:37/3:59 = 1.16). What would this look like for some other times?
3:59 ---> 4:37 (~2:01)
4:10 ---> 4:50 (~2:06:30)
4:20 ---> 5:02 (~2:11:30)
4:30 ---> 5:13.5 (~2:17)
4:40 ---> 5:25 (~2:22)
4:50 ---> 5:37 (~2:27)
5:00 ---> 5:48 (~2:32)
So, asking if Eliud can run 3:59 right away in his marathon buildup / peak is (at least on raw mathematical analogy) equates to asking runners in the parenthetical marathon shape if they can run the corresponding mile time.
To believe that Eliud could *not* run 3:59.0 suggests the existence of 2:32 runners who can't break 5, 2:22 runners who top out at 4:4x in their marathon buildup, 2:17 runners who can't get under 4:30, 2:11:30 above 4:20, and so on.
I have no idea what Chepngetich, Kosgei, etc are capable of in the mile. Ruth opened in 15:11 (4:53.8/mile), so I'm not convinced she couldn't crack 4:30... Connor Mantz just ran 2:08, could he run 4:15 or so? That's 13:10 5k pace, so seems like even in his marathon buildup he wouldn't be out of shot of it.
Could you run 2:07 and fail to be able to break 4:10 on that day? Perhaps.
Now of course Eliud's endurance is likely to be prodigious compared to any mortal, so it's possible these comparisons break down at top end. Still, perhaps people with experience in the 2:10-2:25 range can report some anecdotal data: e.g. Did anyone ever run 2:17 on a day they couldn't crack 4:30?