I am still not sure wether I have mentioned this design idea in the old announcement thread or not.
Regarding the pool situation I think the time has come to think about a sort of distributed pool design which is only coordinated from a central instance.

Most of the amount of work a pool has to do is to verify the received shares to calculate the reslulting deadlines.
Set the fact that miners can do this work instead of the pool there has only to be some sort of verification.

If we look at the pool like a p2p network it is possible to let miners verify the shares for other miners before the pool actually verifies them on its own.
This design requires to add some additional logic into the miners for the pool but may cut down the cpu costs for verifying shares in time to almost 0 since the pool servers would only require to test the submitted share deadlines which have a chance to mine the block (the others get verified by the miners of the pool).

I know this requires some coding and testing but in the end this enables to scale almost as much as bandwidth allows.

To explain this design a bit more in detail you can think of these behaviours as examples:

minerA finds a deadline and submits it to the pool --> the pool dispatches this for verification to 3 random miners --> the random miners decline or confirm and the deadline gets accepted

minerA finds a deadline and submits it to the pool --> the pool dispatches this for verification to 3 random miners --> a timeout (eg. 20 seconds) happens without answer and the pool verifies the deadline on its own

minerA submits a deadline which could find a block --> the pool verifies it on its own and submits the block to the network


the main development which is required would be to design the backend, a protocol to transmit the workloads and the miner integration.

most of this should already exist and only the distributed deadline verification would have to be added.

The nice part of this design is that the miners may also receive a stake of the pool fees for the verification work they do.
I have not played with numbers yet but this means that even small miners can benefit from large miners and it allows to count in even high deadlines (eg. <50m seconds).