Quoting @ssastry steps from irc,
- no req timeouts (current master after the merge); reduce cpu timeouts to a smaller value .. say 2 minutes .. restbase will retry those requests which would fail again
- req timeouts at 110s; they don't remove the cpu timeouts; cpu timeouts at say 2 minutes or a little higher .. restbase will not retry those requests
- req timeouts at 110s; we solve the problem of not having a mechanism to halt the wt2html pipeline; and cpu timeouts exist to deal with the runaway cpu computation scenario