It was even 9235k the 16th. I considered the total output volume too, but it made no sense to me that the thief would need any change, because he has no reason for not spending the entire inputs, except to obscure what he's doing.
OTOH, the volatility of the total output volume was so high those days, that I find it impossible to see an indication for a $1000k move.


Now I have a reason to get more popcorn and hope for a happier ending.

Meanwhile I checked for confirmation of a $1000k steal the 15th of June 2011, but found none.
The estimated transaction volume on that and the following days is about 230k for the entire bitcoin system.
See here.
It is not plausible to me that a thief would risk a reversal of his theft, when all he needs to do is to spend the coins to a new address.

The saying is that about 2000 btc were stolen that day, conjectures about many more stolen coins, e.g. 500k, seem to be widely rejected.
Since the 1000k theft is key to your plot, I wonder if you have a particular source. Anyway, I now like your plot even better, guess why 

I wished I found a flaw in your plot, because it is devastating for anyone's hope of ever getting stuck assets out of Gox.
But so far I didn't find any. Congrats, that all sounds quite reasonable to me.

Looking at the volume charts between mid 2011 and today, Gox may have gained about 200k-300k btc in total from fees.
That lines up well with about 800K btc missing, operating cost can probably be neglected in that calculation.

Keeping the keys for 1000000 btc on the server would not have been necessary. Once there was a substantial amount of btc on hot wallet addresses, they could have sent them to cold storage with one click, even if this was a paper wallet being locked in a bank safe.

What's left, at least for me, is to check for any confirmation of a $1000k steal on that day. We are looking for $1000k coins which mostly entered Gox during the weeks before and then were all spent more or less at once. The thief would have had to do the latter in order to prevent Gox from spending them first using a backup of the wallet, after the theft had been detected.

I wonder how plausible it is that the fact of such an amount of coins being stolen could be kept secret. But never verifying your wallets against the blockchain would  certainly be helpful. In that context, the late detection of of double-spends due to TM is even plausible without introducing TM on purpose.

Realistische Transaktionskosten können nicht höher sein, als die von Konkurrenzsystemen mit gleicher Leistung.
Falls es gelingt, ein Konkurrenzsystem ohne kostenbasiertes PoW zu etablieren, könnte das Bitcoin-PoW nicht mehr finanziert werden, nachdem die Finanzierung durch Ausgabe neuer BTC beendet wurde.

BTC würde m.E. dann nicht mehr funktionieren.