Competitive auditing is very good at a specific class of bug and I think people over-generalise from that. What a contest format finds reliably is the local, mechanical defect. Reentrancy, a missing check, an arithmetic edge, anything a reader can spot inside one function with the contract in front of them. Those are real bugs and finding them cheaply is worth a lot. What it finds badly is everything requiring context the contest cannot supply. Whether the economic assumption behind a parameter holds under stress. Whether the governance path can be captured. Whether this contract composed with that unrelated protocol produces an exploit that neither has alone. Those need someone who understands the business, has weeks, and is willing to say the design is wrong rather than that line 214 is. So the honest positioning is that a contest replaces the first week of an audit, not the audit.