Another spark pool error - Application id is null

A couple things I want to point out are:

- livy ID is available, indicating that the orchestration was initiated, but the pool did not want to begin processing. 

- The "Spark History Server" link is grayed out.  This also means the problem is with the launching of the notebook on the pool, rather than with Apache Spark's core.

- I have many, many other spark bugs at the moment, and I'm at the limit of what I'm allowed to open at Mindtree.  If someone else encounters this, please open a support ticket in the Power BI admin portal.

- the failed notebook was mid-way thru a parallel pipeline that launched about 40 other notebooks.  This was the only one that behaved abnormally.

- My theory is that the (hidden) spark pool/cluster was probably unhappy at this particular moment of time.  It was probably busy doing something that (theoretically) shouldn't imact user workloads, like performing an autoscale operation or something along those lines.

-  There is no way for customers to have direct visibility to monitor their spark pools.  We have visibility on a per-notebook basis but, for the cluster as a whole, we have no way to inspect it for problems.  This limitation will inevitably prevent us from digging deeply into Microsoft bugs, and wont allow us to find our own workarounds either.  I'm convinced that the only path forward is to open a support request (SR) with Mindtree.  I will probably do that once the severity grows to be more than two or three of these failures per week.  (Even if they agree there is a bug, I'm certain Microsoft will demand that customers should implement workarounds, like "retries".  If a failure happens, try again.  If another failure happens, try again.  Eventually the notebook will work, and Microsoft will also get to add more CU's to the meter.  It is a win/win)