I use a slight variation of that: I can't make the thing perfect right now because of unknowns I can turn into knowns only by doing the good enough version, hence trick myself into thinking the "do it" part is part of the "design for the perfect version" part, and generally discover that my initial vision was too broad and that the good enough version I just completed not only is much closer to the goal than I thought but it's more efficient, easily refactorable, and generally nimble, as well as immediately useful. In a way I'm fighting a natural tendency to excessive generalisation with deceitfully self-imposed serendipity.

There is another situation where procrastination kicks in and the above does not work: when I do have some weird, "code smell style" feelings about blind spots in my approach but things need to subconsciously percolate until I get a a-ha moment. In such scenarios I should not work on the thing and do something entirely different as if I try to force myself into it then consciousness and subconsciousness seem to deadlock each other on a shared mutex.