You do raise some valid points, I think on the whole I'm not as picky as you are about the output
I mainly read the books on my Kindle but I also have a Sony PRS-300 (I won it) which I put the odd epub on for my girlfriend. So far I haven't had any real issues with them barring the odd exception.
I also don't do huge bulk imports but rather one to 10 at a time imports because the metadata generally needs tweaking. For the most part as long as the title and author fields are filled out Calibre does a good job of fetching the correct metadata, sometimes there are mulitple entries which requires the user to select the best match. I've yet to import a book that I wasn't able to find metadata for, this could vary dramatically for other users though depending on where they reside and the language of the books they read.
I wouldn't be up to the task of writing my own python scripts to do this unfortunately but I'm glad you pointed out that it would be possible, maybe other people that come across this thread will have better programming skills than me
As I originally stated this would just be a handy feature for some users, for more advanced users who really like to get their hands dirty it would probably lack the fine control they want. For other people it may be great.
Anyway thanks for the feedback, and Kovid (and/or other programmers) if you think this is worth implementing then cool, if not then I appreciate you taking the time to hear me out