Would love to see what sort of message you're getting.
However, one other thing to look at...
Something "new" I learned at SmartBear Connect was the difference between "Run" and "Launch" with regards to TestedApps. Run will wait for the application process to return before it continues. "Launch" will launch the application and then keep going.
So... what it sounds like is that the application is being "Run" but there is some sort of long delay in actually having the process return. So, I'm still curious as to whether or not the process for your application is memory resident after the failing "Run" command. If not, then there's something going wrong with the launching of the application. Perhaps, still, a timing problem that the closing of the application from a previous test is not closing the application fast enough so that, when the next project kicks off, it gets the "Run" command while the process is still resident in memory but then it goes out of memory before it can return the process... Again, something with timing is my guess. What you can try is change the command to "Launch", add code to check to see if the process is present.... if it's not after x number of seconds (while loop), try "Launch" again... repeat a specific number of times until you determine failure and then log an error, stop the project, and move on.