First, on the idea of choosing a big active space:
All of quantum chemistry is about convergence. We can’t practically choose an infinite basis set or all the excitations, so we have to instead find a way to get the relative energies “about right” with a finite basis and only some excitations. This means we have to balance our treatment of static and dynamic correlation. If we pick an active space that’s too big, we start treating dynamic correlation when we meant to just treat our static correlation. We’ll get some correlation “too soon” compared to in a balanced treatment, and our comparisons will be off. So if you want to do a more sophisticated multireference treatment after your CASSCF, there is such a thing as an active space that’s too large!
That brings me to the next point. CASSCF is a multireference method, but it’s the multireference version of Hartree-Fock. Just like Hartree-Fock is a good starting point for CI or coupled cluster methods but not something you’d trust quantitatively, CASSCF is a good starting point for MRCI, CASPT2, or one of the many multireference coupled clusters out there, but not something you should trust quantitatively.
Now for your system, you could use a multireference method, but you also have the option of just going out to high excitation levels. A “multireference problem” can still be captured by single reference methods, but it takes longer than normal to converge. I imagine that by the time you get to CCSDT(Q), you’ll pretty much have your answer. After all, you only have six electrons, and two of those don’t want to leave the core. You can check by looking at how the relative energy changes as you change the correlation level. I’ll warn you that Psi4 can’t do CCSDT(Q). You’d need MRCC or CFOUR for that. However, Psi4 interfaces with both of those, and they’re both free.
If you want to use a post-CASSCF multireference method in Psi4, you can either use Mk-MRCC or something in the Forte plugin. I’m not familiar with Forte, but the manual for that is here.
Good luck!