Fix the race condition bug Jed reported. A time jump on startup could cause

an apparently overly-long (or even negative) ledger interval. The fix is to
start up time synch earlier and to tolerate slight negative ledger intervals.

src/SNTPClient.cpp

@@ -166,7 +166,9 @@ void SNTPClient::processReply()
 	if ((mOffset == -1) || (mOffset == 1)) // small corrections likely do more harm than good
 		mOffset = 0;
 
+#ifndef SNTP_DEBUG
 	if (timev || mOffset)
+#endif
 		Log(lsTRACE) << "SNTP: Offset is " << timev << ", new system offset is " << mOffset;
 }