Also, don't believe everything that prominent members of the Bitcoin community have to say about alternative chains. In particular, I know some people think that the number of confirmations doesn't matter and all that matters is the total expected time of the confirmations, so that 1 10-minute-average confirmation is more secure than 3 3-minute-average confirmations. If you read Satoshi's paper it's clear this isn't true; the number of confirmations is actually more important because transaction security increases exponentially with more confirmations. (His paper has approximate figures; you'll notice that accepting 1 and 2-confirmation transactions is fairly risky.)

Generally, if you have a large amount of bitcoins on a given PC, being extra-cautious about third party software (be it an Alt-coin client or a particularly fancy casual game) is advisable.

You can have more than one wallet dat.

You just need 1 portable bitcoin client that can safely reside on an encrypted and backed up volume, and one regular one for day to day tiny stuff. It's not like you routinely send 5000+ BTC, no?

Incidentally, cobbling together a somewhat workable portable bitcoin is pretty straightforward.

People with only one computer can still securely isolate different wallets & apps from each other by using privilege separation. For example on Linux, run bitcoin/namecoin/i0coin/etc under separate user accounts, and chmod 700 their home directories.

It has long since become a tradition to only patch stuff up when an actual attack emerges, hence leading to emergence of a tradition to demonstrate the "seriousness" of attack.

While I personally would rather follow a different approach, this state of affairs is by no means limited to Bitcoin community or even IT in general, and is here to stay.

I'd like to remind everyone that Gavin, with all due respect, is not an angry   deity and can neither drown people he vaguely disapproves of nor feed them to the lo(l)custs.

One would think that someone who even tries to code knows to rather not say such things to programmers, lol

Although I don't think the fiasco with Namecoin is over and I do believe we have made it hard for BitcoinExpress.  With that said I looked into the block chain rewrite this is real and can be done while lagitamitly mining bitcoins.  This exploit is quite scary for Bitcoins as a person with 20% of the hash rate can write the block chain.

The good news is it can be fixed and Namecoin is a test of such fixes.

I would STRONGLY suggest you look into this Gavin.

It's hard enough to get the half-baked alt chain software to run at all (and speed is of the essence knowing they are all quickly collapsing pyramids) never mind configuring a VM with appropriate hardware access.  Here are steps I've taken which I think are "good enough" to be advice -- it's worked for me for 4 shitcoin chains so far.

1.  Don't use Windows and pre-built .exes.  Just don't.  Ever.  Nothing inherently wrong with Microsoft software, but it is well understood and commonly used by the botnet types.  Staying out of the monoculture is a form of security by obscurity.
2.  Create a new account with no group membership.  I call mine "goatse" for obvious but nostalgic reasons.  Make absolutely sure that account doesn't have read or write access outside of their home directory.   Make doubly sure they can't read the raw hard drive device.
3.  Log out of your main account and into that account whenever compiling or running the alt chain software.  Remember that compilation & installation scripts are code!
4.  Do not browse exchange sites you have coinage in and definitely do not save passwords in the browser when logged in as this account.
5.  If you log into this account via ssh DO NOT enable X proxying.  It's trivial to read your keystrokes, do screen captures, etc when X is proxied.  Let me repeat this one, make sure X proxying is disabled.  If you can type 'xterm' and see it show up on your main account's screen you're vulnerable. 

And yes, I even follow this for official bitcoin software.  On a different account than "goatse" of course.