In terms of usability, I have been setting up a test FME server machine that replicates the production one (Safe has been very nice to give testing licenses that match the number of engines we have on production) and have been running some of the jobs on the test box to verify they work. I'm working on finalizing a FME Best Practices documentation that I'll post up once I get it done. I strongly recommend setting up a replica of each of your production datasets that you write to in a test environment, updating all of your workspaces to write to those test datasets, so you don't muddle with production when writing.

The big things that I've caught include:

- Issues with engines after running a bunch of jobs in succession

- Issues with engines running jobs with different versions of python