It is literally not possible to reach a point where improving skills becomes even comparably difficult to improving root stats.
To open a skill at 5 you need to reach the actual peak of mortal potential, Exp 10, in all that skill's roots. Advancing skills at Exp 1 through 4 requires a small number of routine tests and then their relevant stat's rank in either difficult or challenging tasks, where you would need both types to advance the stat.
Like, let's look at Axe and Lance. Suppose you started a character at B5 Power/B3 Agility, which will root both Axe and Lance at B2, you choose to open Axe at B2 but not Lance, and let's give them a fairly reasonable B4 Will as well.
Per the standard training rules, going to B7 Power/B5 Agility, which will root Lance at B3 when you open it, takes 4 difficult and 2 challenging Agility tests (2/1, 2/1) and 6 difficult and 3 challenging Power tests (3/1, 3/2) in addition to the 7 routine Power tests to open Lance that you can grab at any time. Actual gameplay can be odd and unpredictable, but if we want something like a common in-system benchmark, let's look at the practice regimen you'd need to accomplish this. Agility practice has an interval of 3 months with 2/4/8 hours for routine/difficult/challenging practice, while Power practice has an interval of 1 month with the same time requirements. You'll be pushing to the limits of your will for the first six months or so but it'll get easier after that.
- Months 1-3: one challenging Agility test (8 hours) + 3 difficult Power tests (4 hours)
- Months 4-6: one difficult Agility test (4 hours) + 3 challenging Power tests (8 hours). Your Power improves from 5 to 6 in month 4 and the tests reset.
- Months 7-9: one difficult Agility test (4 hours) + 3 difficult Power tests (4 hours). Your Agility improves from 3 to 4, and your Power improves from 6 to 7.
- Months 10-12: one challenging Agility test (8 hours) + 3 routine Power tests (2 hours).
- Months 13-18: two difficult Agility tests (4 hours) + 4 routine Power tests (2 hours) and some slack time. Your Agility improves from 4 to 5, and Lance opens at B3.
Okay, now what's it take to get Axe from B2 to B3? 2 routine tests and 1 difficult or 1 challenging test. Axe is a martial skill, and martial practice has the same time interval and hour requirements as Power practice does. I'll leave it as a bit of an exercise to the reader, but you can get Axe from B2 all the way to B5 (B3 needs 3 routine and B4 4 routine in addition to your stat choice of difficult/challenging) in literally the time you have spare from this intense training regimen to open Lance 1 point higher.
Or, heck, skip right to the end, after all that intense training's over. I'll just spot you all of that (but the practice to open Lance, anyway). It still takes as much practice (and an arguably similar amount of real-world scenarios) to advance Lance from B2 to B4, assuming you took it at character creation and never used it since, as it would to open it at B3 if you didn't have it. Getting to a position where it's easier to open a skill higher than you could improve one that started lower requires multiple jumps in its root, attaining the limits of mortal ability, or both, and in most of those scenarios the phrase "deliberate neglect" wouldn't be too inappropriate.
Rather than removing perverse incentives, your proposed solution would make them universal.
I can see a couple edge cases in the current system, where you're right on the bubble of both your aptitude and a root increase and you really want to force a big tough stat test somehow, to get over before your training finishes, or similarly when you're only an aristeia away from a stat epiphany and you want to force a stat test to bump your shade before your training finishes. (I say "force" like that's a thing you can even accomplish. Skills at least have known Obs and obvious world hooks for being a thing you're doing. Stat tests kind of putty in the gaps in most cases; you can guarantee a few by scripting specific intricate conflict actions maybe, though you can't really force an intricate conflict your GM isn't up for and there's no guarantee your opposition will roll high enough to count. There are some fixed stat Obs like Health for recovering from wounds, but good luck chasing up an Ob 8 Health check when even a mortal wound is only Ob 6.)
But all you would accomplish by making root changes retroactive to their rooted skills is remove the "edge" from those edge cases. They'd just be cases. Rather than being conditional on specific circumstances, it would always be good to try and force stat tests and it would always be good to save all your artha for stat tests.