We can add iterative sampling of multiple correlated fields per shell (e.g. matter and convergence, even if that has all of the known issues) without any fundamental changes to the current algorithm: Instead of generating field1, field2, field3, field4, ... we simply generate tuples of (field1, field2), (field3, field4), .... This only requires some changes to the correlation length, and adding a new size parameter to the generator function.
We can add iterative sampling of multiple correlated fields per shell (e.g. matter and convergence, even if that has all of the known issues) without any fundamental changes to the current algorithm: Instead of generating
field1, field2, field3, field4, ...we simply generate tuples of(field1, field2), (field3, field4), .... This only requires some changes to the correlation length, and adding a newsizeparameter to the generator function.