Thank you, Eodw1, for the file.
My first point would be—to address that you need to think in smaller parts of the objects. This object is basically (in its start-shape) just a sixth of the needed shape. Mirrored and multiplied creates the shape. But you can of course work with one object from the start. It is sometimes not clear what workflow might cause more time. But as a rule of thumb, analysis the shape first, then decide about the procedure.
After that, my way would always be, to think of a constant check if the Subdivision Surface (HyperNURBS) has sufficient polygons, or (!) way too much. The more polygons you use, the more it turns any further step perhaps into huge amount of work. What the Subdivision Surface does and how it works is a longer story, and needs to be explored. Without a good feeling and experience of that option, modeling is not funny.
To optimize the mesh, first create a “nice” mesh from the Loft, nice means here, try to get an evenly structured mesh. In the Loft switch off caps and adjust the Spline in the Intermediate-points. After “Current State to Object (CSTO)—with “Weld” you might get ride of some too small triangles.
My suggestion, set the amount of polygons and the subdivision of the source spline as low as possible for that. The Loft (NURBS) should be set only to the cap that is needed.
Checking the Subdivision Surface (SS) always, will show you that the current model might create already an unusable shape. I assume that you use the SS, as it might become quite a challenge otherwise to get the shape look nice.
One of the main things to avoid in SS modeling is N-Gons and as far as possible triangles. N-Gons are great during modeling, but need to be solved into quads (four sided polygons). Yes, some exceptions exists, but a clean model pays of always. QUADS, and as close to a square as possible. With that in mind, the underlaying mesh can be quite rough for the work with the SS.
If there are triangles, then this needs a support seam, to create a distance to the triangles. Which works only very well, if the triangles are flat to the support. As a rule of thumb, SS uses the geometry like a rubber and pulls a surface out of it. Lots of points and closer to each other pull more than others. Hence the artifacts.
After the Spherize is applied drag the first supportive frame via Extrude. Based on the Spherize this will match the surface nicely, perhaps do it twice and the second time just a tiny little bit. After that the CSTO is used, then Select>Ouline and Extrude work starts. First to get some support “frames” around the shape and pull it down via Y move. The Extrude must be tiny, very tiny!
The steps need to be monitored perhaps with SS on and set to Catmull-Clark, SS level to 1 in this case. The Extrude shouldn’t be set to Subdivision nor N-Gons for this.
After the first small drag down, do another super tiny Extrude, and pull again, so you get to the shape of the black shell. Then again tiny, pull down and now set the Coordinates Y-size to zero and adjust the Y position where the feet should end. Delete the polygons where no feet are. In the image the foot area is selected. Notice the rough mesh in th other screenshot—no SS.
I have not accounted here for any details on the center area, which might be great to address while in spherize mode to pre-sort some of the polygons already for later events.
I hope that brings you closer to your target.
All the best
Sassi