Cranktrain wroteit doesn't matter if there are no vertex transforms happening, just having a mesh of 5 vertexs attached to a bone is going to be a worse case than a four vertex region, even if there's no deformation of that mesh.
This doesn't quite make sense. A vertex transform is when the local position of a mesh vertex is used to compute the world position. If you have a 4 vertex mesh or 4 vertex region (all regions are 4 vertices), then you are doing 4 vertex transforms. Fewer vertex transforms is better only if computing the vertex transforms is a bottleneck. Eg, if you are spending 100ms on applying timelines and 1ms on vertex transforms, cutting your vertex transforms in half isn't going to make any difference. Note the difference between 4 and 5 vertex transforms will never, ever by a performance problem.
For mesh deformation, when a timeline is applied, a value is stored for each vertex which is used to offset the vertex when doing vertex transforms.
The performance impact of how many pixels you are drawing (fill rate) mostly depends on your video hardware. If this is your bottleneck, you simply need to draw fewer pixels.
The number of draw calls you are doing is separate from the fill rate. This can be a problem if you are switching slot blend modes or don't have all your images on the same texture. Note using PMA atlas images can prevent draw calls when changing between normal and additive.