Google crawl errors being caused by JavaScript in our CMS?

Googlebot attempts to follow 'links' which only appear in the CMS script.

[dkchristy]

About 2 1/2 months ago, we started seeing crawl errors across all of our sites in Google Webmaster Tools. The phantom 'pages' all followed the same pattern:

mysite.com/1234567/examplename/

After some investigation, we realized Google could only be pulling these strings from one location - a single line of embedded JavaScript created by our content management system. Here's an example of what the offending line looks like (this is fake code, just to help you visualize):

flmanageCM.fsrollup_mlc="/2621551/billybob/" + location.hostname + "/" + flmanageCM.fsrollup_default_pc;

In this faux example, the resulting error page would be mysite.com/2621551/billybob/. Here's where it gets weird: after about a month, all of these crawl errors started disappearing from Webmaster Tools - we thought we were off the hook. This week as the sites are being crawled, they're all coming back. So:

-Has anyone else seen a situation where Google has tried to crawl strings that look like 'folders' or 'pages' in JavaScript?
-Anyone observed recent changes in Google's protocol in handling JavaScript?

Thanks for any thoughts or observations!

[phranque]

i think this has been common behavior for quite some time.

you can obfuscate your links in the javascript by url-encoding all reserved characters.