[allthesethings]

The problem with making guesses about any of the technical decisions is that this is a software project with hundreds of people in multiple studios. There are no design efficiencies to be had here: somewhere within those hundreds of developers a rationale for taking on a more complicated architecture arises - a specific feature, or combination of features, that causes a cascade of requirements inflation.

For example, one reason they might want higher precision is for editing and processing persistent data: a need that doesn't apparently have anything to do with the physics simulation, but subsequently ripples down into "well, if we want to allow editing in real-time, then..."

It is interesting work, but nothing that you would want to emulate as a best practice for making a Better Game. It's just a result of the software aspects being scoped and executed in such an uncompromising way. A good majority of the work to date has been on the organization-building work of teams and processes that can execute at this scale.

[everyone]

What??