A vPC peer link outage takes down single-attached servers only

During night maintenance, a contractor replacing optics on LEAF2 pulled both peer link cables (Eth1/53 and Eth1/54) at the same time. The peer link was down for seven minutes. LEAF1 is vPC primary, LEAF2 is secondary, and the peer keepalive runs over mgmt0. Both leaves route VLANs 10 and 30 with SVIs and HSRP, and have routed uplinks to CORE1. Dual-homed servers such as SRV1 (vPC 20) saw a short blip. The backup server SRV2, single-attached to LEAF2 Eth1/10, and a licensing appliance on LEAF2 Eth1/11 were unreachable for the whole seven minutes, although neither port ever went down. You are leading the post-incident review.