Find where reality first diverges from expectation.

The most useful point in any failure is where things first stopped matching what you expected. Everything after that is downstream and mostly noise. Finding that first divergence narrows the search enormously, because whatever went wrong went wrong at or just before that moment.

When tracing a problem, work backwards to the earliest point where reality and expectation still agreed, then move forward slowly from there. Ask what you assumed would happen at each step. The break is usually somewhere much earlier than where the trouble became visible.