This is the second time you've suggested that the Bitcoin reference code is responsible for your robbery.   I inquired about this claim before and I don't believe I got a reply: https://bitcointalk.org/index.php?topic=81045.msg899922#msg899922  Luke-jr also expressed skepticism: https://bitcointalk.org/index.php?topic=81045.msg899911#msg899911

 I fail to see how any system which has private keys for online realtime 'hot wallet' usage could be defended against an attacker which has root access to the selfsame systems.   Even if you used a multisignature wallet and machines inside separate security domains an attacker with that level of access could simply impersonate the web application's legitimate withdraws.

That said— if there is some flaw or omission in the reference client which could make high value installations more secure all the developers would love to hear about it.

What I am reasonably confident of is that while you're quite possibly smarter and have more time on your hands than any one of the people developing the publicly available reference software, you're not smarter than all of them combined.  ... And a bug that sends 18kBTC into a black hole (as MTGOX's custom code did with a few thousand BTC) is no better than having code stolen.  

There are significant advantages in working with a larger user base to test out and harden code before putting it on mission critical systems, and those advantages almost certainly outweigh the many troubles and limitations in the reference client.   Moreover, many aspects of Bitcoin security require that you be a part of the majority clique— even if the majority is "wrong"—, if you can be moved onto a minority chain you can be robbed.   Because the significant super-majority of the network (users and miners) are using the reference client, its critical that any client be bug for bug compatible with the block rejection rules in the reference client or be at increased risk.  So it very much is in your own interest to invest resources in improving the publicly available software than reinventing the wheel.