Re #7215. Change testing strategy to avoid occasional failures.

I think the sporadic failures of this test were because alg2 had not
started up even after the sleep. Calling validateInputs() instead
accomplishes the same testing goal and avoid having to execute this
instance of the algorithm at all.

Code/Mantid/Framework/DataHandling/test/MonitorLiveDataTest.h

@@ -48,11 +48,11 @@ class MonitorLiveDataTest : public CxxTest::TestSuite
 
 
   /** Create but don't start a MonitorLiveData thread */
-  IAlgorithm_sptr makeAlgo(std::string output, std::string accumWS="",
+  boost::shared_ptr<MonitorLiveData> makeAlgo(std::string output, std::string accumWS="",
       std::string AccumulationMethod="Replace",
       std::string EndRunBehavior="Restart", std::string UpdateEvery="1")
   {
-    IAlgorithm_sptr alg = AlgorithmManager::Instance().create("MonitorLiveData", -1, false);
+    auto alg = boost::dynamic_pointer_cast<MonitorLiveData>(AlgorithmManager::Instance().create("MonitorLiveData", -1, false));
     alg->setPropertyValue("Instrument", "TestDataListener");
     alg->setPropertyValue("UpdateEvery", UpdateEvery);
     alg->setPropertyValue("AccumulationMethod", AccumulationMethod);
@@ -122,11 +122,7 @@ class MonitorLiveDataTest : public CxxTest::TestSuite
 
     // This algorithm if OK because the other is not still running
     IAlgorithm_sptr alg2 = makeAlgo("fake1");
-    Poco::ActiveResult<bool> res2 = alg2->executeAsync();
-    Poco::Thread::sleep(100); // give it some time to start
-    TS_ASSERT( alg2->isRunning() );
-    alg2->cancel();
-    res2.wait(10000);
+    TSM_ASSERT("validateInputs should give the all clear (an empty map)", alg2->validateInputs().empty() );
   }