nexus: allow instance updates to complete when one switch is down - #10389

control: job assigned to worker 01KR1Q68F4TDNQJMFG6YGHCJDW [factory aws, i-0e14e511d4d2b9a5f] (queued for 44 s)
...
task: starting task 0: "setup"
task: process exited: duration 4167 ms, exit code 0
...
task: starting task 1: "rust-toolchain"
task: process exited: duration 15869 ms, exit code 0
...
task: starting task 2: "authentication"
task: process exited: duration 18 ms, exit code 0
...
task: starting task 3: "clone repository"
task: process exited: duration 10195 ms, exit code 0
...
task: starting task 4: "build"
...
| a out/probe.tar.gz 3037K
| a out/propolis-server.tar.gz 38727K
| a out/pumpkind-gz.tar 26647K
| a out/switch-softnpu.tar.gz 392863K
| 
| real     1:18.945246797
| user     1:16.310309517
| sys         4.376252489
| trap        0.006918955
| tflt        0.000000000
| dflt        0.000000000
| kflt        0.000000000
| lock        0.000000000
| slp      1:16.208117100
| lat         0.983988441
| stop        0.000221580
task: process exited: duration 879757 ms, exit code 0
|W| found 1 output files
|W| uploading: /work/package.tar.gz (2431118960 bytes)
|W| uploaded: /work/package.tar.gz