It's not any different than having 153 steps in any other computational sense. Even limiting ourselves to elementary arithmetic, horrendous opaqueness arises with 153 operations spanning the whole set. Are we going to pretend like arithmetic is a systemically problematic because of this? Any non-trivial formal construct is potentially dangerous.

If you're having trouble reasoning about how variables are unified, it's either because you never actually built a strong gut intuition for it, or it's because you're writing Very Bad Code with major structural issues that just so happen to live in the type system. In this case it's the latter. For an HM type system, 153 choice points for an expression is ludicrous unless you're doing heavy HKT/HOM metaprogramming. The type system, and more broadly unification, is a system to solve constraints. Explosive choice indicates a major logical fault, and most probably someone naively trying to use a structural type system like a nominal one and/or a bit too much unsound metaprogramming.

Thankfully of course, you can simply just specify the type and tell the compiler exactly what it should be using. But that's not really resolving the issue, the code still sucks at the end of the day.

Now higher order unification? That's an entirely different matter.