The "field tests" in our test suite are very static in the parameters they use, and therefore some bugs may fall through the cracks, as was pointed out by @bobverity during the review process.
I propose that field tests have a few variant runs, and not always the same run.
In order not to overload the CI server, I would suggest we use set up several different runs for each wrapped program and randomize which parameter set gets run on each CI run.
Thoughts?
This as addressed instead by thoroughly improving the test suite for the external program wrappers, which was done during the latest changes to the code (compartmentalizing and testing each unit).
The "field tests" in our test suite are very static in the parameters they use, and therefore some bugs may fall through the cracks, as was pointed out by @bobverity during the review process. I propose that field tests have a few variant runs, and not always the same run. In order not to overload the CI server, I would suggest we use set up several different runs for each wrapped program and randomize which parameter set gets run on each CI run. Thoughts?