It [0] happened only once, but I do not think it is specific to 2c 1518B and E822Cq. Interestingly, the ~1% loss on DUT was accompanied with ~1% unsent packets. Port 1 was the one seeing smaller RX and TX, and I doubt this is duration stretching, looks more like NIC/driver bug. The testcase itself is low priority, opening this (as opposed to other seemingly infra issues) because subsequent testcases are not affected, and the symptom might happen later in more important VPP test cases.

[0] https://s3-logs.fd.io/vex-yul-rot-jenkins-1/csit-dpdk-perf-report-iterative-2410-3n-snr/4/log.html.gz#s1-s1-s1-s1-t5-k2-k5-k14