To simulate randomization, would you be able to add a new selector looking for some attribute text that may or may not exist, and delay it by say 1000ms? Such text string would depend on the Webpage.
The CSS selctor might be looking for "003" in any the ID attribute for any division, or some HREF that ends in "a.htm", such as:
div[id*='003']
a[href$='a.htm']
This selector should be a sibling of other valid selectors, so that the scrape doesn't get broken if this text is not found in the attributes. The DIV above, if it exists, should contain short text, so as not to clutter up the results.
If you set several of these "contingent" text selectors with different delays, would that simulate some randomization?
I haven't tested this. Just an idea that has been bugging me too.
Edit: I am not sure whether a null result will still spend the same delay time. Perhaps adding a child under that "contingent" element (perhaps a link or "element" selector) will actually create that delay.