How do I fix intermittent hardware problems I can't reproduce?

You can fix most hardware issues now, so you know your stuff. The intermittent ones beat you because they hide, working fine the moment you look, then failing when you turn away. That is genuinely one of the hardest kinds of problem there is. Struggling with it does not mean you have hit your limit. The trick with intermittent faults is patience and evidence. Since you cannot force the failure, you catch it instead, logging, monitoring, and noting the exact conditions when it does happen. Over time the pattern shows itself: it fails when it heats up, when it is jostled, under a certain load. The clue is almost always in the circumstances, not the moment. Next time one of these appears, resist the urge to fix it on the spot. Instead set up a way to record what is happening when it fails, and gather a few instances before you act. Let the pattern come to you. Intermittent problems yield to observation more than to force.