By the time 60 or 90 minutes has gone by since the last block, the mempool would be so large with fees that 100% of hashpower should be online trying to get them
Conversely, after a block is mined the mempool should be empty and without a subsidy it may not be profitable to mine until mempool reaches a certain size.So there will be pressure for low hash rate before 10 minutes and high hash rate after 10 minutes, giving a much more consistent 10 minute block time.
But is it profitable to buy hardware, knowing that it will only run some of the time, when the mempool gets extra full? Maybe. Depends on both the hardware cost and the variable costs.In any case, I agree it will be more profitable to mine when the mempool is bigger, but the *degree* to which miners will smooth over block-time variability depends on their costs.