The CAN status variables are all explained in can.h:
So these are simply indications of RX buffer 0 overflows. These are no "real" overflows, as RX buffer 1 was still available, so the frames were not lost and the rxbuf_overflow counter was not incremented.
The driver could clear this condition to avoid signaling this as a CAN error. I think we kept it that way because an RXB0 overflow already indicates a lot of traffic on the bus. This could also result in a warning log, but the CAN framework doesn't know how to distinguish driver specific errors from warnings.
Going back to your earlier advice, I think you're asking us to monitor the rate of can errors in the log, and observe if that changes with different bus timing.
And also monitor can status. With my current firmware (existing timing, single sampling, both buses in active mode) I see can1 is a bit busier than can2 (10 million vs 8.6 million). Can1 has had one genuine receive overflow.
It's quite possible there are others from earlier in the day, when I had debug level on. I've now set the level to warn so it should be easy to spot everything significant. I'll update when I have something to report.
Chris