There is now a DDOS. Presumably orchestrated by someone trying to exploit the GOX bug in order to drive the price lower.
For a long time, Bitcoin protocol allowed transactions to be expressed in multiple different ways, yielding potentially several different Hashes /ID numbers for the same tx.
They fixed this many months ago so that the qt wallet and the command line client only ever issue tx in a standard 'canonical' format that has a single possible hash/ID, and the standard Bitcoin daemon won't forward or mine tx that have one of the noncanonical representations.
But Gox didn't fix their own software, and every so often Gox would issue a tx that didn't have the canonical form. There were a couple of miners that still allowed these transactions to go into blocks, so the noncanonical transactions would be delayed while all the miners that didn't allow it got blocks, then finally go through when they got "lucky" and hit a miner that hadn't updated software.
A couple weeks ago, after many warnings, the Bitcoin network reached a threshold and the full nodes stopped accepting blocks that contain noncanonical transactions. Miners that hadn't updated their software yet started getting orphaned blocks every time they included one of them, which happened about as often as Gox emitted them. The bug at Gox bit their asses as payments made from Gox having a noncanonical representation were refused by the network.
Keep in mind, at this point this is less than one one thousandth of the payments coming out of Gox. Essentially it only happened in some small fraction of the tx where Gox was spending its own unconfirmed change txouts.
But as Gox started spinning on that little dildo, some malefactor went, "Hey, I bet we can give them a much bigger dildo to spin on! We can make Bitcoins cheap enough to buy a bunch of them if we make this look like a bigger problem!" And they monkeyed up a little script, which they then pushed out to some botnet somewhere.
Their script takes regular transactions off the Bitcoin network and re-emits them in noncanonical form, in an effort to absolutely melt down anybody who hasn't completely fixed this bug yet. Gox goes from having an occasional problem where it thinks it's made a payment that it hasn't actually made (because it made it WRONG), to being absolutely helplessly reamed on a giant monster cock. The few remaining miners that haven't updated their software go from getting one extra orphan every once in a while to getting every last block orphaned. Meanwhile the rest of the network is spammed with thousands of bogus transactions a minute, all of them copies/alternate expressions of 'real' transactions. The real transactions (in canonical form) get accepted by the network and the bogus ones don't, but the network is under strain.
This sudden flood of noncanonical transactions results in exposing an edge case in the qt and cli client software, as well as the software used by *MOST* exchanges. This edge case is benign, but confusing. The problem is that the noncanonical transactions can show up as unconfirmed tx for a while in your client. They will never confirm, and the correct transactions that they are copies of will confirm eventually, so this is essentially harmless. But it can make it appear such that you've accidentally made a payment twice, or appear that a payment coming to you has been sent twice - one of these payments will eventually confirm, and the other will not, but for a few hours, it will look like there is something wrong. And the payments have different ID numbers.
Bitstamp now looks at the situation, going, "hmm, there's a DDOS underway and I bet a bunch of our customers are going to be making transactions to try to "correct" this double payment situation which doesn't really exist, and when they understand what actually happened, and that there were no actual double payments being made, they'll have made transactions they wish they hadn't. We should head off that situation somehow...." and so they suspend trading because they figure that's easier than trying to explain everything to their user base.
Other places, like Coinbase, carry right on trading because they figure if their customers make transactions because their customers are confused, then that is their customers' problem not theirs, and besides they're in a more litigous jurisdiction and afraid that if they suspend trading that will open them up for lawsuits, either for failing to execute or for contributing to a panic.
Meanwhile, Bitcoin developers go 'doh!' and get to work fixing it so that noncanonical transactions which won't ever confirm also don't ever show up as unconfirmed.
That seems very plausible, I can see all the actors involve make the decisions that you are guessing they made, that is still probably not exactly what happens but mtgox is using an customised wallet, there has been a DDos attack
What is your explanation concerning the 102$ sell on btc-e?