A CertiK audit just claimed 15.52M TPS on a live mainnet. What would it actually take to verify that?

A CertiK verification report is claiming a mainnet blockchain hit 15.52 million transactions per second under live conditions. It wasn't a test/benchmark environment but in the mainnet itself.

That number is so far outside and inflated to the normal range for any other blockchain that it's hard to accept at face value even with a 3rd party audit. CertiK is a legitimate auditor but imo and audit is only as good as its methodology, and massive numbers like this require scrutiny of exactly what was measured.

Before anyone dismisses this as impossible or treats it as settled, I want to understand what independent verification would actually look like. Three questions I'd want answered:

• What is the precise definition of 'transaction' being counted? Economic transfers and lower-level state operations are very different things and conflating them is a classic TPS inflation technique.
• Was this measured under synthetic load designed to hit maximum throughput, or under conditions representative of real network activity?
• The audit references a bare metal, no-VM architecture as the explanation for the number. Is that explanation technically coherent, or is it a post-hoc justification?

If anyone has actually read the methodology section of the report rather than just the headline number, I'd like to compare notes and discuss. I'm going to dig into this one and will post what I find.

I'm not trying to "debunk" it. Just doing the work before forming a view. Thanks.