You seem to be again fundamentally misunderstanding what it means to run incompatible versions of software. It doesn't matter what you think rational miners will do. Once the 750 of 1000 blocks are found, there is no going back. It only takes a modicum of hashing power to start publishing blocks that are incompatible with 1MB nodes. This is not about "pushing Blockstream to get their act together." It's about avoiding the risk of breaking bitcoin forever.
and that "modicum of hashing power to start publishing blocks that are incompatible with 1MB nodes" wont be accepted by 1mb blockers and 1mb blockers wont stale their attempts. they would carry on hashing their own blocks and make blocks a few seconds later.. eventually even if it takes several blocks when 1mb gain height, they cause the C big blocks to orphan off..
Yes, 1MB nodes won't recognize a > 1MB block as valid. It doesn't follow that 1MB miners will make blocks "a few seconds later," especially if they have a relative minority of hashing power. Again, if you are suggesting that 70% majority of hashing power cannot find 3 blocks in a row, you are completely wrong. Here is an example from tonight:
397332 1 hour 4 minutes 1987 49,928.62 BTC F2Pool 976.54
397331 1 hour 22 minutes 554 6,649.97 BTC F2Pool 452.19
397330 1 hour 26 minutes 590 18,539.01 BTC F2Pool 341.15
F2Pool, with 27% of the network's hashing power, found 3 blocks in a row. During this time, there were no valid
orphaned blocks. So not even a stale block found by another pool. (i.e. Group C found 3 blocks in the time that Groups A and B found 0)
Again, there is nothing that will cause Group C to orphan its previous blocks if they are valid blocks built on the longest valid chain. They may break Group A and B's consensus rules, but since they do not break the consensus rules of Group C, they are valid based on Group C's node software and will no blocks will be orphaned.
miners wont risk it at 70%.. yes the setting will be active but miners wont push for more than 1mb at such low levels as the orphan risk is still apparent.
they would wait for a higher number.. and just treat the 2mb setting as an unused buffer for the future. when they are comfortable,
and when that time comes (im guessing 90%). then and only then would the small miners not be able to catch up to cause orphans and the small miners should upgrade or be left behind.
which they should have done earlier as they had enough warning
If [rational] miners won't risk 70%, why the hell are we activating new consensus rules at 75%? You keep talking about miners like a single entity that you can predict. Again, it doesn't matter that you believe miners will use it as an "unused buffer" (whatever the hell that means). Gavin's code activates the rule change at 75% of mined blocks; after that, any CPU contributing hashing power can determine whether there is a chain fork based on the node software it runs:
You can't talk about "miners" as a single entity. A node is either running one version of the software or the other, assuming they are incompatible (in this case, the are). That means that after the hypothetical 28 days, if 70% are running node software that accepts > 1MB blocks, once any single miner or pool publishes a block that is valid based on 2mb parameters but not 1mb, we have passed the point of no return. "They won't push the envelope?" How could you pretend to predict the actions of every single CPU contributing hashpower to the network?
You've already predicted that 0% out of 100% of hashing power will publish a block breaking the old consensus rules -- that the new limit "is nothing more then a buffer," even if 70% ran the node software at some point in order to activate the new rules. On its face, that is extremely unlikely given that it only takes one actor with a modicum of hashing power to cause the node software of a majority of miners to enforce the new consensus rules.
So once Gavin's 28 days are up and any one miner or pool publishes a >1MB block, hashpower ceases to be the question at all. The question becomes which chain node operators consider valid.
Do you then go on to predict that 100% of nodes will be running one version of the software (1mb limit) or the other (2mb limit)? Because if not, we will inevitably have an irreparable chain fork.
If you want to know whether a hard fork activating with 70% of hashing power can break bitcoin into multiple blockchains (presumably forever, as too much value will have changed hands to conceivably "roll back")... the answer is unequivocally YES.