I mean, okay, if you can redefine rationality to be completely subjective, then sure, rationality is flargikriggendurf.
EDIT: This is similar to any other sort of debt that people find themselves in. I weighed 220lbs, couldn't bench 100lbs (maybe 2 reps), and could barely run 1/4 mile at a 12-minute mile pace before wanting to pass out. So I was in a health/fitness-debt. Optimally, I could've set aside 2 hours a day to some combination of aerobic and strength training to achieve my goals in a short period of time. But that required a massive change of habits that was unlikely to stick. Instead, I ran 3 days a week (well, walk/ran at the start), learned to set aside my time for fitness activities, and improved my cardio. As I managed my time more effectively, I added in other fitness activities that got me to my strength goals as well. Adding it all at once would have (for me) been overwhelming and I likely would have failed (again) to make a routine of it. Adding in each part piece-by-piece was the far better strategy, and entirely rational, bypassing the normal anxieties of my mind by making the large change occur over several smaller changes.
Optimizing the likelihood of consistently following through with a plan is very rational.
Humans are not merely "irrational", they positively behave and think in very specific ways. No human is immune to this. It is, therefore, perfectly rational to discuss rational ways of managing the specific human behaviors that lead to irrational behaviors.
This is also while ritual denunciations of "homo economicus" are really quite sophomoric... the nonexistence of rational economic actors does not mean that you can simply model humans in whatever way you please to make your preconceived notions work. It means that you've got to go learn what humans actually do, which is going to be very challenging, and take that into account, which is almost certainly going to shock anyone who tries it. It means the problem is suddenly immensely harder, not easier.
Admittedly, this is because "homo economicus" is sophomoric in the first place: it proposes an actor who has no computational or informational bounds and sees the world only in terms of a single real-valued variable called "utility", which is generally constrained only to be any monotonic, continuous function of their monetary net worth. This actor then optimizes expected utility with no regards to modeling causal structure or to which observables are actually ergodic.
No such creature ever has, or in fact ever can, exist in the real world. In limited domains where ergodicity holds, computational needs are few, and small sample sizes can yield very good inferences, we can do thought experiments about what such a creature would do, and act according to those if we please. But holding that such behavior is the normatively correct way to act when it's actually impossible is the kind of religious thinking one gets when composing "normative theories" without experimental basis.
Logic depends on a subjective choice of axioms, and an unstated one is "a plan that motivates permanent improvment is better than a temporary one".
Option A (snowball): Total Cost if successful: $X Likelihood of success: 90% (just making stuff up here)
Option B : Total Cost if successful: $Y Likelihood of success: 80% (again, making stuff up)
If 80% * Y + cost of failure > 90% * X plus cost of failure, even if X > Y, then a rational actor will choose Option A.
Sure, the "chance of success" is subjective, but ignoring it doesn't make the likelihood of failure go away.