Currently there is no option to apply symmetry during 3D classification without alignments.
In most cases this isn’t necessary, but in certain cases it is useful. Specifically, when dealing with a symmetric system and looking for a subset of particles that have improved resolution, enforcing symmetry during classification can sometimes give better results, so it would be useful to have (perhaps with the option of relaxing symmetry?).
Depends on the situation but in certain cases yes (and it is also usually faster not having to deal with an n-fold larger particle set).
They are also distinctly different approaches, as with symmetry expansion the symmetry of the reference volume is not enforced (which can lead to classification on small asymmetric features that are not relevant to the investigation, which can in turn require the use of more classes to resolve the feature of intrerest).
We have had cases where symmetry enforcement during classification without alignments in relion allowed us to identify a state which we could not pick up otherwise, and I imagine the same will apply in certain cases in cryoSPARC.