Follow-up ticket to:
Each space view should track all wgpu errors in a separate error tracker. This way we can detect whether a space view failed to render, typically resulting in a frozen or black screen on wgpu-core. Once we do, we can display / overlay an error message on the affected space view only and remove it once we recovered again.
Note that experiments indicate that on Dawn / Chrome WebGPU we're less luckily: Instead of failing a single command buffer, we're failing the entire frame resulting in a black screen. Mitigating this should be possible (e.g. stop rendering broken space views until there's a reason to resume!) but might not be as trivial.