I'm also a programmer, and I'm totally agree with your - there will be more bugs.

Bugs here are the same as MtGox hacking - nothing special.

The only way to overcome it - to split the money so no single address could take all the money so fast. Some "money flow cap" will do.

Again, look at sophisticated cold wallet systems on centralized exchanges. For example, OKcoin claims they have only 2 humans with access to multisig cold wallet in different locations on the Earth.

So just another exchange hacked, not too big deal for the whole Ethereum ecosystem.

I think hardfork will be certainly bad idea. Nobody hardforked after mtgox.

But they could do softfork and exchange eth stolen for 1% bounty. That is better. The Attacker will try to go in the court... It will be fun to see the process. If he win, then Ethereum computer will win its creators in the court :))

So overall I think LONG DAO, LONG ETH now. :)

One way to reduce number of bugs is pure functional programming language, like Haskell
As far as I could understand from http://vessenes.com/deconstructing-thedao-attack-a-brief-code-tour the problem is either typo Transfer instead of transfer or malicious typo

So git annotate of dao code could lead the light. Not sure if all the commits history is available.