The mere act of enumerating IE7 incompatibilities in these applications (let alone fixing them) represents a development cost that some companies will put off for as long as possible.
Then you could upgrade $(Initech) to IE7/Firefox, and slap the proxy between the users and the legacy web apps, and everybody's happy. When the internal developers upgrade the legacy web app, they can either keep it on the un-IE6-proxy, or upgrade its output and lose the proxy. IE6 can go away, we can stop wasting time with IE6 compatibility, everybody's happy, world peace, the end.
Disclaimer: I don't understand the nature of IE6 bugs very well at all, and I don't know if this is turing-complete or AI-complete or what.
"we got a problem here, uh yeah, the website is down"