I'm using NMKD Stable Diffusion on my end paired with DreamShaper checkpoint.
I've been having decent success unconscious/death poses but mostly because I'm using image input and just using pictures of someone laying down and then going back and adding in prompts to adjust the face. If you're going from text prompts alone, it might help to say, lying down and then get the variation you like, and then go back and use that image and modify the expression.
A combination of image input, inpainting and some editing(both minor/major) has gone a long way for me. For editing, it could be a matter of rotating the hand a certain way and then painting over the original position with colors from the background to let AI fill in the blank/blend it in.
I haven't delved too much into action poses just yet but it might do well to try something like "battle/fighting stance" and the like to see if that yields anything more promising. Same thing with weapons and items, although I imagine using some editing to tweak the general shape of the weapon or even editing them in might fare better. I say edit them in but you still have to run it through AI so that it properly fits/blends into the rest of the image using inpainting.
I've had decent success with blood by just doing crude drawings of blood splatters, stains and blood dripping from bodies and letting AI interpret it paired with "blood".
Though my character design hasn't been as intricate as yours, I've started with a naked person and then gone back and just did skintight bodysuits if anything as I was more practicing with the scene.
