A middle way could be to only observe robots.txt for crawling, and not for displaying pages. So once a page is grabbed, it's available forever. But if a page is covered by a robots.txt exclusion, it won't be crawled.