Normally use them, keeping all non-dynamic objects in spatially relevant sublevels so they can marked as static and therefore reloading times are faster. If you have 1 person only touching levels then potentially lumping it all in the persistant will be ok (apart from being a bit sloppy workflow wise) but if any more than 1 person uses a level at a time you’re potentially creating bottle necks for your team. Personally I would never do that, I’ve worked on enough AAA titles to discourage a workflow like that even if I’m only working on a personal prototype - you have to keep your workflow trim and team oriented if you want to be remain part of a team.
So we generally use sub levels for
distinct ENV levels for efficient memory and performance
distinct BP levels for blueprints and dynamic objects like destructibles, pickups and level-design stuff (these roughly align to ENV levels)
NAV meshes (so dynamic nav mesh rebuilding can be switched off by hiding the sub levels and also AI guys and other level designers can adjust nav-relevant objects in isolation)
Cutscenes - so cinematics team can do their thing without interruption or bottlenecks
Same with sounds
So it depends on your team, task divisions, memory and performance concerns, how you’ve been taught in regards to workflows and keeping everything neat, tidy and compartmentalised.