How important or realistic is the recommendation to keep under 30 nodes?
Has anyone had more than 30 nodes and ran into major issues?
How does memory usage impact this?
Discussion posts and replies are publicly visible
It is a recommendation and not a hard rule but I strongly suggest you follow it.
But splitting up your processes it will not only help with performance but it is easier to maintain as well for the development (making modifications in process designer to a large process is difficult). It also makes it easier to refactor code as it favours reusability and finally helps with debugging/troubleshooting in production. Having in long flight processes can be difficult to manage over time.
See here for more details:
Agreed with Mathieu - in general there are almost always some kind of logical grouping that you can use to combine similar nodes into a sub-process. To be honest, the big concern isn't necessarily memory if you have short-lived processes, but it's just much easier to maintain, test, and update with smaller process models.