Don't focus to much on these numbers. What's interesting in those numbers in the opportunity to have a relativily large samples of usecases. And different usecase bring their own set of issues. Problems are interestig because you have to solve them. This is precisely what I want to talk about.
We do not impose strong constraints on uploaded projects So we can have many differents situations
In order to scale you will need to replicate your environment That mean replicating all projects in memory Also Problems of synchronization of loaded projects in the different units
Address the problem of sacalabiliy and deployemnt
Need change our point of view about projects.
Adapt your services to the project's execution context
Projects may be very differents They cannot be managed the same way
We deal with large typology projects there is obviously no a single strategy