The new drivers installed on the DAQ yesterday as part of WP4583 did not improve performance as expected, and in fact made it worse, so I have rolled back the changes to h1nds1, h1fw1, and h1broadcast0. I have left h1dc0 as it is since the change there does not seem to have had any ill effects. It's not clear to me why the same change showed an improvement on the DAQ test stand, other than the traffic level is significantly higher on the production system and must influence the driver behavior or the system configurations have diverged enough to affect the results of testing of this nature. Here are the downtimes for reboots:
h1nds1: 15:40:35 - 15:45:54 UTC
h1fw1: 15:48:23 - 15:52:30 UTC Frames starting/ending in this period will be missing - use h1nds0/h1fw0 for data in this gap.
h1broadcast0: 16:00:40 - 16:02:40 UTC