QA also works best for features that the user can meaningfully interact with. The more esoteric, interconnected, or complex the failure is, the more it gets muddled with other reports and weird conditions (it breaks on 3pm when the moon is in equinox and I have my mouse pointed north). That’s an over exaggeration but the sentiment is accurate. QA will frequently lack a deeper understanding of how the software works and is connected. That ignorance can be valuable (finds blind spots) but has trade offs (assumptions are made based on imperfect observation of cause and effect where human biases lead you down dark hallways). Speaking from experience with lots of manual QA which is actually a rarity these days.

The other thing to consider is, if you’re careful, user feedback can be obtained more quickly. If you can keep things stable, then you won’t upset your users too badly and you’ll avoid the need for QA to act as a stand in.