Example run: [0].
Only one test on one testbed is affected (at least from those run in daily job), only one NIC (100Ge2P1E810Cq), only one driver (AVF). It is the test that could (in theory) produce the largest TX burst from VPP.
Of course, the actual number of packets is not negative, so this bug is maybe in some TRex component. The reported number is close to (negative) 2^32.
Also this only happens in the first of 10 MRR trials. In daily jobs it happens regularly, in verify jobs it happens only sometimes or it needs specific conditions to trigger (but never happens on TB37).

[0] https://s3-logs.fd.io/vex-yul-rot-jenkins-1/csit-vpp-perf-mrr-daily-master-3n-icx/214/log.html.gz#s1-s1-s1-s5-s3-t3-k2-k9-k8-k13-k1-k2