I ran the mutect2_pon workflow (a GATK best practice workflow) for creating a mutect2 panel of normals. I ran the workflow on a set of normal samples from the TCGA THCA cohort. The workflow failed. Using the FireCloud GUI, I tried to drill down and investigate the reason for the failure. The submission launched a single workflow, when I attempted to look at the call data for that one workflow, the GUI says its loading workflow details, sits there for several minutes, and then shows nothing. I did look on the bucket and saw the workflow resulted in a large scatter operation on the mutect2 task (387 shards from what I could tell). I didn't feel like going through all of those shards to see which of them was the source of the failure.
So there are two problems:
1. I could not run a best practice GATK4 workflow on TCGA data (and the failed run cost $30!)
2. I could not use the FireCloud UI to investigate the reason for the failure
The first problem should probably be reported to the GATK forum, but before doing that the second problem needs to be resolved.