The problem
When a "crawl" or "search" robot runs, it visits many pages one after another and collects data from each one. But all of that collected data is only saved to the database after the entire job finishes.

What "done" looks like
Start a crawl or search job on a site with many pages.
Kill the job partway through (or simulate a timeout/error).
The results gathered before the failure should still be visible/available in the database - not lost.