Erik, I think you're missing the point - code that doesn't work is useless. The argument of the code being 'maintainable' so we can easily fix it doesn't gell. What if the maintainable code has the wrong architecture? If it doesn't work it is useless.

Have you looked at the code in you car's fuel injection computer? It is probably unmaintainable -I've disassembled a few in my time. Nevertheless the code must be doing its job as you drive your car everyday without a glitch(I assume this is the case). So if my car had a switch on the dashboard selecting between 'nice code but doesn't work' or 'code that works' , I think I know which one I'd be selecting! How about the code in the games for the Atari 2600 (flashback time kids!) - it was probably the tightest code one could imagine. Maintainable - nah, it just had to work.

Put simply, when I write code, I write it to work and I write it to be maintainable where possible. If I have to squash a zillion features into 2k of flash, then I won't be using 'c' and I'll probably break all the rules of maintainable code to fit it all in. Obviously, I try to avoid such things! Such is the case of embedded systems.

But, rather than labour the point as you have failed to convince no more than one of us of your argument, I think we'll stop right here.