Seen this on 2 arm64 hosts (dazzle and starmie-kernel) for hirsute:linux(generic-64k) 5.11.0-38.42 for sru-20210927. This did pass before on dazzle, so might be a regression but only for the generic-64k flavour...
10:42:23 DEBUG| [stdout] # IPv4 runtime torture
10:42:23 DEBUG| [stdout] # --------------------
10:47:23 DEBUG| [stdout] # TEST: IPv4 torture test [ OK ]
10:47:23 DEBUG| [stdout] # ./fib_nexthops.sh: line 186: 63568 Killed ipv4_del_add_loop1
10:47:23 DEBUG| [stdout] # ./fib_nexthops.sh: line 186: 63569 Killed ipv4_grp_replace_loop
10:47:23 DEBUG| [stdout] # ./fib_nexthops.sh: line 186: 63571 Killed ip netns exec me ping -f 172.16.101.1 > /dev/null 2>&1
10:47:23 DEBUG| [stdout] # ./fib_nexthops.sh: line 186: 63573 Killed ip netns exec me ping -f 172.16.101.2 > /dev/null 2>&1
10:47:23 DEBUG| [stdout] # ./fib_nexthops.sh: line 186: 63574 Killed ip netns exec me mausezahn veth1 -B 172.16.101.2 -A 172.16.1.1 -c 0 -t tcp "dp=1-1023, flags=syn" > /dev/null 2>&1
10:47:24 DEBUG| [stdout] #
10:47:24 DEBUG| [stdout] # IPv6
10:47:24 DEBUG| [stdout] # ----------------------
10:47:24 DEBUG| [stdout] # TEST: Create nexthop with id, gw, dev [ OK ]
10:47:24 DEBUG| [stdout] # TEST: Get nexthop by id [ OK ]
10:47:24 DEBUG| [stdout] # TEST: Delete nexthop by id [ OK ]
10:47:24 DEBUG| [stdout] # TEST: Create nexthop - gw only [ OK ]
10:47:24 DEBUG| [stdout] # TEST: Create nexthop - invalid gw+dev combination [ OK ]
10:47:24 DEBUG| [stdout] # TEST: Create nexthop - gw+dev and onlink [ OK ]
10:47:24 DEBUG| [stdout] # TEST: Nexthops removed on admin down [ OK ]
10:47:24 DEBUG| [stdout] #
10:47:24 DEBUG| [stdout] # IPv6 groups functional
10:47:24 DEBUG| [stdout] # ----------------------
10:47:24 DEBUG| [stdout] # TEST: Create nexthop group with single nexthop [ OK ]
10:47:24 DEBUG| [stdout] # TEST: Get nexthop group by id [ OK ]
10:47:24 DEBUG| [stdout] # TEST: Delete nexthop group by id [ OK ]
10:47:24 DEBUG| [stdout] # TEST: Nexthop group with multiple nexthops [ OK ]
10:47:24 DEBUG| [stdout] # TEST: Nexthop group updated when entry is deleted [ OK ]
10:47:24 DEBUG| [stdout] # TEST: Nexthop group with weighted nexthops [ OK ]
10:47:24 DEBUG| [stdout] # TEST: Weighted nexthop group updated when entry is deleted [ OK ]
10:47:24 DEBUG| [stdout] # TEST: Nexthops in groups removed on admin down [ OK ]
10:47:24 DEBUG| [stdout] # TEST: Multiple groups with same nexthop [ OK ]
10:47:24 DEBUG| [stdout] # TEST: Nexthops in group removed on admin down - mixed group [ OK ]
10:47:24 DEBUG| [stdout] # TEST: Nexthop group can not have a group as an entry [ OK ]
10:47:24 DEBUG| [stdout] # TEST: Nexthop group with a blackhole entry [ OK ]
10:47:24 DEBUG| [stdout] # TEST: Nexthop group can not have a blackhole and another nexthop [ OK ]
10:47:24 DEBUG| [stdout] #
10:47:24 DEBUG| [stdout] # IPv6 functional runtime
10:47:24 DEBUG| [stdout] # -----------------------
10:47:24 DEBUG| [stdout] # TEST: Route add [ OK ]
10:47:24 DEBUG| [stdout] # TEST: Route delete [ OK ]
10:47:25 DEBUG| [stdout] # TEST: Ping with nexthop [FAIL]
Update: the fact that the torture tests reports a lot of "killed" is at least no regression. And I just saw amd64-generic reporting this test failed on a run for ADT (though a different test and for IPv4). So this might just be coincidentally showing up with the same failure here...