As per WP 13504 we made some changes on h1daqdc0 to see if they would impact the error rate*.  We turned on the irqbalance daemon to see if that would help with IO load.  We are going to let this run a day or two and evaluate what things look like.  We are looking at other ideas (timing changes, network config changes, possible code changes to lock cores and network interrupts together better) but are going to hold off until the control room is ready.

* In the daqd, the CRC errors are a misnomer.  It really means late data, not corrupt data.