only log all Req details for problematic requests

[Dieterbe]

not tested yet

[Dieterbe]

tested it by inserting the following in the node.Post code, and comparing traces. it behaves as expected

+       rem := time.Now().Unix() % 10
+       if rem == 0 || rem == 1 || rem == 2 {
+               time.Sleep(11 * time.Second)
+       }
+

one thing worth noting is that with the cap at 45, the trace is only about 15kB, whereas we are allowed 65kB, so we cut a bit too aggressively. but i think it's ok, logging the first 45 entries should be a good tradeoff between usefulness and disk space usage.

[replay]

LGTM