If we stored a bitfield recording all partitions with "recovering" power, we could avoid loading any partitions for proofs that neither recover power nor mark sectors as faulty. This would reduce the cost of a window post to:
load actor, miner info, deadlines, current deadline (4).
store everything except the miner info (3).
Loads/stores to save the proof (~1-3).
We'd expect less than 6 stores, less than 7 loads, and no sends.
If we stored a bitfield recording all partitions with "recovering" power, we could avoid loading any partitions for proofs that neither recover power nor mark sectors as faulty. This would reduce the cost of a window post to:
We'd expect less than 6 stores, less than 7 loads, and no sends.