Interesting, I was thinking of code to analyze this, who knew miners would pop out like this when graphed! I also am pulling my own stats out with bitcointools, but at a block a second, it takes a while. My stats add nonce, time, a spent status, and # of transactions to the table of extranonces, but I don't thing I will bother with writing my own analysis code.
http://bitslog.files.wordpress.com/2013/04/all10-5.jpg (image is wacky so just a link)
I see an interesting aspect, that the extranonce of the non-spending satoshi miner (if it is correct that red=spent and black=unspent) is increasing about four times as fast as others - it's steeper block finding slope indicates that it is
significantly lower at block finding than it should be. Remember, computers are not competing against each other during this whole run, they are competing against the difficulty.
I believe this is because this miner is using different code than the released Bitcoin. It may have a getwork-style mining, where multiple CPUs are mining through it, and part of this increases the nonce faster than the block finding rate of the released Bitcoin. The nonce may increase faster if Satoshi was discarding fast block finds to make the block rate more regular (which they are statistically - pull out and analyse one series of increasing extranonces). Something that we see at about 1/10th, and again at 5/8ths of the chart, that the slope changes (along with a daily block rate change), I think that there was some change made in its code or the number of miners configured, maybe Satoshi added something that made the miner inefficient for few weeks and fixed it, or tweaked code to discard any blocks faster than a "heartbeat".
There is one scenario where I think this makes sense: Satoshi created a special "network-supporting" miner. This special always-on miner with it's own code base may have never created permanent keys or had a wallet - in that NOTHING appears spent from it's mining.
Without an always-on miner for people to try out the currency and keep confirmations rolling (and another node for them to connect to), the money would just plain not work. This miner runs different code than the release others are running, evidenced by the identical slopes of everyone else. This may be the "x" IRC username connecting over Tor - it would listen for IP addresses to connect to but normal bitcoins couldn't get it's IP address from IRC.
Another thing interesting is that you can see other black lines in the graph, other people mined for a while with release code and have never spent their generates. We know some people have early 50BTCs, but we can also surmise a few of these lines may be lost coins.
What you see are different THREADS of the same PC. Possibly two or four threads were running concurrently.
Each thread has it own ExtraNonce, but all threads are started almost at the same time.
There is no evidence to support this. In fact this disagrees with your discovery that most mining was done by one rolling extranonce. It may have been Bitcoin #1 + xMiner, as referenced in my previous posting thread. Remember, Bitcoin 0.1.0 doesn't mine unless it has a node connected to it - it counts the nodes, and if the number is 0 or has changed to 0, mining is disabled.
The Satoshi super miner might not have started until Satoshi's normal Bitcoin or another regular Bitcoin user was also connected. Satoshi may have a wallet with his own coins, but never will use the backbone miner's coins. New miner at #64 may be Hal getting Bitcoin working.