In other words, while rendering a single 2000 sample OpenEXR Multilayer image, I open another instance of Blender, with no scene and a clean compositor. After that I drag and drop a low sample image previously done of the same render, with the same Cryptomatte pass in it.
Possibly I can start compositing and export the crypto-matte from here, applying it to the 2000 sample image after it finishes. Is it right? I see the edge of the matte untouched, as Cryptomatte works with object mesh, maintaining the luma black and white matte clean (as an object buffer, I think).
I don't know if I'm explaining it right, but I think that the title is quite clear.