Rendering differences help

What's the difference between render cache fusion output and render cache color output? My workflow is basically ( Generate optimized media > Set render cache to user > Turn on render cache color output > Set render cache fusion output to on ) or simply just ( Generate optimized media > Set render cache to smart ). I don't fully understand what I am doing so can someone shed some light on this? What should be the most appropriate scenario to use render cache fusion output and / or render cache color output?

There is a section in the Introduction of the Reference Manual called Improving
Performance, Proxies, and the Render Cache. I haven't read the whole thing, but that seems a logical place to start.