The shim process is mostly iddle waiting for instructions from c8d.
This means that the shim process is io bound rather than cpu bound.
In this case, an async workflow would be more suitable for the shim.