Not all failures should cause retries; an HTTP 404, for example, should not. A network failure should. Part of this ticket is figuring out exactly what kind of failures should allow retry.
+;; cases: +;; - silently drop requests +;; - can't resolve hostname +;; - timeout on DNS +;; - can't contact host +;; - connection denied + +;; constant problem: allow it to fail the first time, fix the problem, then check it worked +;; +;; intermittent problem: fail the first invocation only; check the first sync +;; worked through it. + +;; all of these should sync correctly after the condition is resolved, or sync eventually through the flakyness condition + +;; - single record fetch fails => the rest of the sync should work