What problem does this solve or what need does it fill?

Users should be able to use the full extent of the wgpu draw APIs. While users have access to the wgpu::RenderPass via the RenderContext provided to Node::run, the RenderCommand API only has access to a TrackedRenderPass, which has a limited subset of draw APIs exposed; in particular, all of the multi_draw... variants are missing.

What solution would you like?

Either add more wrapper methods to TrackedRenderPass or find some other abstraction that makes it feasible to work directly with the wgpu::RenderPass from the RenderCommand trait.

What alternative(s) have you considered?

Implementing my own Node instead of using RenderCommand. This would be a major change.