The presence of a persistent arb premium was what made me back off Gox. As a quant trading guy it just seemed too fishy that a one-way arb could persist. I wrote an arb engine to take advantage of the price difference but I concluded there was too much credit risk in it. Which turned out to be correct. At best the arb was explicable as slow operations getting money out of the bank, but that would in the very best case be total incompetence. Plus the forums were saying a lot of not nice things about the management, so definitely failed the sniff test. Remember it's not the balance of probabilities that matters; it's the worst case.

Coinbase, by contrast, has VC backing that you would think means they'd find someone who understood regulatory issues. They'd also have links to proper tech people who understood security and exchange coding, a pretty small subset of coders since it's quite specialised. They may not have started with everything required but chances are they've found the money to buy it by now.