Daniel Larimar incorrectly
claims in that video that it is not reliable to validate transactions in parallel multithreaded. Nonsense.
Why nonsense, it depends on linearity of the system. For a linear system order doesn't matter, for a non-linear one it does.
PS: We assume that multithreaded execution can't ensure a specific order of events, which is pretty reasonable for current architectures without placing a lot of memory barriers which would degrade the performance significantly.
Because (as indicated/implied by my prior post) it is more sane to design your system holistically such that ordering of transactions is an exceptional event, and not a routine one.
Conflation of "order book" with TX/s is a category error. It is not even clear if a decentralized "order book" can or should have a deterministic ordering, because determinism may allow the market to be gamed. In any case, it is not relevant to the issue of rate of processing TX/s for signed transactions. Separation-of-concerns is a fundamental principle of engineering.