NIC input errors have traditionally indicated problems at the link level (crc errors, runts, etc). Large companies tend to build monitoring infrastructure around such errors in order to monitor for bad network ports. When L3/L4 checksum errors are included in the category of input errors, it breaks such monitoring, as these errors can originate with hardware anywhere on the internet (not even under the control of the local administrator), and do not necessarily indicate faulty local network hardware.
In afb1aa4e6df245d38fd2ba683fa521d5dabe8392, L3/L4 checksum errors were included with input errors in ixgbe, and thus broke our (Netflix) monitoring infrastructure, due to elevated counts on some servers due to IP / TCP checksum errors caused by faulty equipment outside our control. I had patched around this locally, but due to a merge error, we recently lost that patch and just experienced the problem again. So I'd like to float the idea of removing L3/L4 checksum errors from input errors from ixgbe upstream.