The current lotus scheduler implementation has certain downsides that does not allow a storage providers to get the max. out of the hardware they are running. Having a general implementation that works on top of resource availability is great and should be kept. A flexible scheduler should aim to amend the current implementation and allow for storage providers to replace it with their own scheduler logic.
An example of this being planned/implemented: filecoin-project/devgrants#445
Issues:
To be filled in by @f8-ptrk