blocks full again with TPS unacceptably high --> unconf tx set ~ 3000, more than double normal. pools taking defensive action. when will Blockstream devs do something?:
Why do you call this "defensive" action?
when Chinese pools face a series of what they consider to be "large blocks", which in this case means full blocks as you can see from the data, they automatically switch to mining "header" blocks with 0 tx's during the time it takes to process all the tx's in the preceding large block. this is b/c large blocks take a bit more time to process and check all the signatures so the argument goes that they can't afford to waste that precious time so for defensive purposed they just go ahead and start hashing the next block with only the "header" that in this case contain no tx's that might have included an invalid input from the preceding large block.
How about: When they see a new block, they immediately start mining based on that, with an empty transaction list. When they have removed the transactions in the last block from their own list of transactions to include, they mine with the new list. If a block is found quickly, it will be the empty block, else it will be a block with the new list.
You can see that the zero transaction blocks always come a short time after the previous. But they should not need minutes to establish a new list, so I am not sure.
Maybe some miner can chime in and say what really is going on.