Excuse me but what are those images supposed to do?
What's the point of meshing many bones that are joint together?

Maybe he meant the fact that less empty alpha space is better when you pack the images, not inside spine. Meshes are used to deform images, you use them if you're going to deform them otherwise that many points are a burden as well if they do nothing (:
Although there's a technique that exploits meshes to save up on packing space I think you shouldn't focus on it this way given your images, rather, if you want to save up on space:

• try to position your source images so that they occupy less space possible: eg: placing them along the vertical/horizontal line rather than oblique ( this -> | rather than this -> / )
• try to reuse images, for example in your second image, you could just have had a bone, a circle and the other shape, bone placed vertically, and you would later duplicate them and rotate them inside spine, that's saving up on space (:
  Hope it helps!

Spine does not support holes in meshes. This simplifies a number of things and is rarely needed.

Using meshes to avoid drawing blank space saves you some fill rate. The trade off is that the more vertices you use, the more CPU you use. In both cases the differences are very small. While they can add up in an application that is doing a lot on suboptimal hardware, I would encourage you not to optimize aggressively before you actually have a performance problem because chances are you won't have a performance problem.

We would love to fix the crash. Can you reproduce the crash every time? If so, can you post or email a project file (images aren't needed) along with instructions on what steps to take to cause the crash? contact@esotericsoftware.com