so in the first few rounds. they start with empty block.. and start adding more transactions per round.
(much more efficient this way to start with 0tx and add a few every extranonce round)
I'm not sure that's correct.
If they added a few transactions every time they incremented the extraNonce, then we would expect not just to see the occasional empty block, but to also see a not insignificant number of blocks which could have been 100% full but were only 5%, 10%, 15%, 20%, etc., full. Instead, what we actually see (when the mempool is full enough to support it) is blocks which are consistently 100% full, and then the occasional empty block. Surely this suggests that miners just attempt to mine an empty block until they have built their full candidate block, and then switch to trying to mine their candidate block. They can obviously update their candidate block over the coming minutes as new transactions come in to their mempool, but it doesn't look like they create multiple candidate blocks on the way to creating their first full candidate block.
..i remember some time ago someone looked at all the blocks that appeared to be solved quickly after a previous block, and seen not only were there empty blocks but also partially filed blocks.
Are there? I don't remember seeing many such blocks, but I could well be wrong. Can you provide a link?