Well, I've just tried caching the complete HTML of the table and that works fine.... I think we will try it this way.

But it would definitely be interesting to find out the reason for this bug and how to work around it...