Looking at logs, I do not see any deviation before VPP stops seeing the NIC in [0], as opposed to last good run [1].
RDMA core tests still see [2] the same NIC with matching PCI and MAC addresses.
More investigation needed to decide if this is an infra, VPP or CSIT bug (and whether testbeds other than TB27 are also affected).

[0] https://s3-logs.fd.io/vex-yul-rot-jenkins-1/csit-vpp-perf-mrr-daily-master-2n-clx/1317/log.html.gz#s1-s1-s1-s5-s7-t1-k2-k5-k4
[1] https://s3-logs.fd.io/vex-yul-rot-jenkins-1/csit-vpp-perf-mrr-daily-master-2n-clx/1313/log.html.gz#s1-s1-s1-s5-s7-t1-k2-k5-k4
[2] https://s3-logs.fd.io/vex-yul-rot-jenkins-1/csit-vpp-perf-mrr-daily-master-2n-clx/1317/log.html.gz#s1-s1-s1-s5-s13-t1-k2-k6-k1-k1-k1-k1-k2-k13-k1-k1