Sorry, forget all above.
The long run times were Friday night, 1-Jan-2016, the second night the programs were run under COBOL Server. When I checked the run times for last (Monday) night, 4-Jan-2016, I found them to be about the same as their Net Express counterparts! I have gone back and checked the run times for the first (Thursday) night, 31-Dec-2015, under COBOL Server and they were also about the same as under Net Express.
So apparently something extraneous was going on with the server (backups maybe? patching?) on Friday night during our scheduled times and that is what skewed the times.
So for now I guess all is good.
And thank you for priority info, we'll keep that in mind.