Thanks for the report. I will look into the problem.
Normally the `ncr' should only attach 810 and 825 non-A in releng_4. This
works since old-PCI code based driver as `ncr' returned a priority of
-1000 to the new-bus code on successful `probe', and `sym' returned -2000
for the 810 and 825 non-A but returned 0 for later chips.
Given your report, it seems that both the `ncr' driver and the trick that
allowed `sym' to attach the SYM53C876 have been broken in the snapshot.
I will check that ASAP and let know freebsd-gnats-submit about. :-)