My personal experience is different. OT tends to be quadratic in terms of computational complexity and codebase size. As long as you consider a simple case, 1^2 is 1. Then, add more data types (not just text), faulty servers, faulty network, non trivial network architectures... All these problems tend to interact. They easily go combinatorial, but again 1! is 1. I worked with an OT codebase that had C++ templates within C defines and I edited them using regexes (x types of operations => x^2 types of potential transformations).

OT aside, data sync in general is not a solved problem. Maybe I am especially unlucky, but I have yet to see a chat app that has no sync glitches. A chat is not the most difficult data structure to sync, but I saw bugs in Slack, Telegram, Instagram, Gchat, Skype, others... WhatsApp does not even try to sync...

A lot depends on the diversity of your use cases and the number of the nines you expect from the system...