In fact , it's on their best interest to mine empty blocks as long as subsidy exist . By best interest i mean their chances of finding a block is probably higher with an empty template .
... no? The time it takes to calculate the merkle root is minimum. It is really against their interest to not collect transaction fees.
You are correct . My bad on that .
when they see the last block solved. they do not send a message to all asics to halt hashing their current(now stale) block
instead the pool does send them a new header with the just solved blocks hash as prev-block for asics to hash on. and yes its an empty block
the pool manager then confirmed utxo as spent and removes them from utxoset, then adds the new confirmed outputs into utxoset and removes the transactions from mempool
then it creates a template including transactions from mempool that are still unspent that were still waiting in mempool
the pool manager then updates the asics requests for new nonce-extra nonce ranges with the new template
so pools dont wait around to make a new template filled with transactions they set the asics initially to work empty block and then update asics once the blocktemplate is ready