I've been messing around for over week with using combining aspects of loras. Sort of crazy that I have about 1.5k of images genned just from me experimenting with things.
It turns out that the majority of the visual styles is determined by the output 1 block, with the bulk of the image composition coming from the output 0 block and input blocks (the effect getting greater the closer they are to the middle block). The middle 1 block seems to alter the finer details of the image.

The differences in styles does get negated by adding expressions and poses, and going from one artist's style to another isn't that compelling, but it may be useful for a concept lora. You can match the pose and composition a lot more closely if you also use a negative strength for one implementing the style on output 0 and the inputs, but it feels difficult to dial in to the point you may have an easier time with image to image.
I was getting annoyed trying to use a hyperballs lora where it kept turning gigantic balls into a separate ball-shaped object and having the bulge between the legs turn into huge balls.