Yeah I think that will be necessary to limit the cashu mint bandwidth after a block is found. Otherwise miners will all try to redeem gigabytes of ecash at the same time. I think a sliding scale of fees for denomination neatly solves this problem without locking out smaller miners. Smaller eHash tokens pay slightly higher fees, larger denominations pay less. This fee pays for the pool/mint bandwidth. From the miner perspective, optimal strategy is to swap eHash for the largest denomination before the next block is found. This can be made automatic in the miner/wallet client.

For coinbase UTXOs, I think you just specify a NUT-11 locking pubkey and set a boolean flag indicating you want an on-chain redemption. The eHash can only be redeemed by the pool with a signature from this pubkey, proving ownership. The pool will add a coinbase output paying to this pubkey in the next block template it sends out. There will be some edge cases where a slightly oldish block template gets mined, just make up the difference with eHash. Dealing with these edge cases will be the majority of the development work.

Not sure if we want to enable miners to BYO template. Probably not supported initially. But I think as long as the coinbase matches and the transaction fees are within acceptable bounds it should be doable. That's a v2 problem though.