Here's a possible explanation as to the "technical" issues that MtGox has been suffering.
http://www.reddit.com/r/Bitcoin/comments/1x93tf/some_irc_chatter_about_what_is_going_on_at_mtgox/In short, Mt Gox wrote their own custom wallet software, which only worked because of a bug in early versions of bitcoin. The bug was fixed in a recent release. This would cause transactions sent from Gox to fail.
A group of people, thinking that they were being helpful set up relay servers that would take the transactions from Gox, fix them to work around the bug, and then rebroadcast the fixed transactions. The transactions could reuse the same signature (as they were the same transaction), but would be allocated a new TxID.
A second bug in Gox's wallet meant that their wallet client only checked it's own TxIDs, and never scanned the blockchain for other TxIDs on its coins. As a result, it is possible that the wallet ran dry, but the actual wallet client didn't notice, and transactions would block because of attempted double-spend.
Yeah, even though Gox will have a hard time regaining the trust lost over the past months, I'm at 70-80% that this issue is indeed mostly a technical one. It seems plausible - though the resulting chaos might turn a stumble into a fall.
Also the "simple" explanation seems to be that transactions with multiple inputs are more likely to run into a double-spend and not being relayed to the blockchain. It's kinda obvious if this is what's going on - when a 1
BTC withdrawal from Gox draws from 50 addresses it is much more likely that the withdrawal will fail due to one of the inputs already used on the blockchain. This doesn't mean Gox doesn't have 1
BTC - it just means Gox doesn't have 1
BTC on the address it thought it did.
I just don't get why they haven't unforked their crappy spinoff yet. It'll cost them millions in bad PR.
When under pressure it's always better to use something tested as the core and build on top of that. In IS development it is ALWAYS better and cheaper to do it right the first time around.