See bug 934310: There is a mozharness timeout after 3600 seconds, presumably before Talos has given up on firefox. I suspect firefox is hung but the Talos timeout is greater than the mozharness timeout, so firefox is not killed by Talos and no crash report is generated.

The sessionrestore test has been using a timeout of 1000000 seconds -- far longer than mozharness will wait for the harness to complete. As a result, when the browser hangs, the harness does not timeout before mozharness kills it, resulting in a timeout failure (reported in bug 934310) without a crash report. This reduces the timeout to 900 seconds, so that the Talos harness will wait only 15 minutes for each browser session. https://treeherder.mozilla.org/#/jobs?repo=try&revision=e29a7eb9f8298acc6b1fd15eead8d45147336f54 reproduces the hang from bug 934310 with the 900 s timeout and successfully produces crash reports.

Comment on attachment 8884038 [details] [diff] [review] use a 900 second browser timeout Review of attachment 8884038 [details] [diff] [review]: ----------------------------------------------------------------- Good catch!

Pushed by gbrown@mozilla.com: https://hg.mozilla.org/integration/mozilla-inbound/rev/c0b7b3cc327a Use a reasonable browser timeout in sessionrestore talos test; r=ahal