We were noticing some CMS detections weren't working simply because we weren't getting responses from the site we were checking. After some digging around and debugging we came up with two solutions. The first was simple:  add a second server to increase our redundancy. The second solution was to fix a few http protocol bug/issues we found while we were debugging. We just implemented these changes so we don't have any significant numbers about improved detection rates yet but we're confident they will both make our service more reliable.