Test performance regression in f4239876 #9064
Description
Using #8289, I noticed a test duration spike around f423987.
TOTAL TIME TESTS TEST SUITES ENVIRONMENT COMMIT
2853.62s 2138 106 aws d92c7789 [nexus] allow abbreviating ereport 'class' as 'k' (#8804)
2987.08s 2138 106 aws dfac975a [nexus] Add address lot view method. (#8752)
2881.43s 2138 106 aws d8f30c18 [reconfigurator] database support for PendingMgsUpdate for host phase 1 (#8791)
2822.59s 2138 106 aws f4239876 implement Nexus quiesce (sagas, db activity) for upgrade (#8740)
# ^ slowdown of ~250 seconds here
2561.84s 2136 106 aws 0abb3f11 Update Rust crate qorb to 0.4.1 (#8442)
2623.82s 2136 106 aws 78021f0c [nexus-mgs-updates] Add support for host phase 1 updates (#8737)
2542.02s 2132 106 aws c097d37a Update Rust crate diesel to 2.2.12 (#8666)
2561.77s 2132 106 aws cf35d2a8 Return all LLDP configs for each switch port settings object (#8665)
2594.25s 2132 106 aws 1591e218 [reconfigurator] RoT planner support (#8421)
2506.89s 2131 106 aws 6a93f8c4 ci: fix published name for incorporation.p5p (#8824)
2540.78s 2131 106 aws 27feea55 Bump crucible and propolis revs to latest (#8809)
This increases the overall duration for tests by ~10% - it's probably worth investigating why this was so costly, and if it could be less expensive.