Brilliant idea...  don't have a clue how it could be done based on what is
readily available, but one could easily enough write a program to download
ALL the messages into a database and then add up how many times each word
appeared.  Take that list and discard the top common words like "the"
"vanagon" "etc..."  but even then I'm not sure that would get you what you
want...

The "best" solution would probably just be to post a poll or take a survey,
then use those results to query the archives and develop some sort of
meta-FAQ linking to relevant posts, or just authoring a concise answer...

The other thought would be to manually comb through the last one year of
posts and then "assume" that it is representative of multiple years.  It's
not like tires were only talked about in 2003 :)