Shrink The Problem Until It Cannot Hide Anywhere

The smallest situation that still misbehaves is worth more than any amount of staring at the large one.

When something fails inside a large piece of work, most of that work is innocent. Your job is to establish which part. Cut the situation in half. Does the trouble survive? If yes, the fault is in the half you kept. Halve again. A handful of these cuts reduces an enormous search to a few candidates, and the arithmetic works in your favour even when your intuition does not. This method needs no insight and cares nothing for how clever you are feeling, which is precisely why it is dependable.

Reduce the input alongside the instructions. Fewer items, simpler values, shorter text. Keep cutting for as long as the problem persists, and when a cut makes it disappear, put that piece back, because you have just learned something specific about what the problem requires in order to occur. The end product is a tiny example that fails reliably, and such an example is close to a solution because there is nowhere left for the cause to conceal itself.

The related discipline is checking your assumption rather than your instructions. When you are stuck, there is almost always something you have not verified because you were sure of it. That the information arrived. That the branch was entered. That the value is what you believe. Certainty is exactly the marker of an unexamined belief, so make a habit of listing what you are sure of and then confirming the least likely item on the list. That is where it usually is.