If you’re sticking with regular subdivision sculpting then that is certainly the workflow you want. This will let you control the larger forms by working with as few verts as possible, letting you use the subdivision smoothing algorithim to create nice curves.
The same goes if you’re doing something like sculpting from a dynameshed sphere; I usually recommend keeping the resolution as low as possible, updating it only when you pull out new forms, and turning it off once you no longer need it. This will leave you with a low-resolution basemesh so that you can fall back on using subdivisions to create the finer details.
There are some exceptions, and using dynamesh to fuse parts together can be one of them (especially if there is a difference in density between the two parts, or you want the result to have nice curves like a sphere or cylinder). Dynamesh doesn’t smooth the model in the same way subdivision does. If you dynamesh an 8-sided cylinder then the result will still look like an 8-sided cylinder regardless of how many points it now has. You might be able to smooth it back into a nice curvy shape using some smooth/polish/etc brushes, but generally its easier just to make the starting shapes as smooth as you want the dynamesh result to be.
If there’s a lot more work to be done on the fused forms then I would follow up the dynamesh with a zremesh pass in order to once more have a lower resolution model to subdivide and sculpt on.
