Empty response 444 in Nginx confuses search bots

This causes false alarms, for example in the Google Webmaster Tools stats it will show many errors logged (correctly) as connection errors - the bot thinks your site is offline or partially/temporarily broken etc. As we deny access for some URLs bots should never attempt to crawl anyway, we should respond with standard 403 error instead, so the bot will understand it as "nothing to see here!" and give up, instead of "let's try again, they are probably offline".

For reference, here is the full list of URLs by default denied for all known bots, as either they are never cached or are known to cause serious system overload when crawled (often to index hundreds of empty pages, as it is the case with calendar module) or are just not expected to be crawled and indexed:

location ^~ /search
location ~* /(?:autocomplete|ajax|ahah)/
location ^~ /admin
location ^~ /audio/download
location ~* (?:cgi-bin|vti-bin|wp-content)
location ~* (?:calendar|event|validation|aggregator|vote_up_down|captcha)
location ~* ^/(?:.*/)?(?:user|cart|checkout|logout|flag)
location ~* /(?:node/[0-9]+/edit|node/add|comment/reply|approve|users)