I tried FrankenPHP on our (externally heavy I/O) workload and it wasn't worth it
I spent a few hours playing around with FrankenPHP on a fleet of servers that run as a microservice, their sole purpose was getting a request in, making external calls to 3rd party APIs/websites, and returning the result (the kind of microservice that should really be rewritten in something like Golang, but that's for later).
I wanted to test if we could speed things up by moving from PHP-FPM to FrankenPHP, but after intensive testing, it isn't worth it for us. I could've predicted this, but wanted to test it anyway :D
tl;dr: if you only spend < 1% of the PHP lifecycle bootstrapping the framework, and the other 99% waiting on external I/O (whether that's database/cache/api), you're not going to see the gains.
More details are up on my blog: https://ma.ttias.be/laravel-octane-vs-php-fpm-lessons-learned/