Many tests flaky after #8603

After commit cc1aba6a4f3649bc0ed699de9199456dd6fd434f (#8603), running large batch of tests with high parallelism (e.g., cargo nextest run --no-fail-fast on a 32 core machine) often produces many failures (50-90, depending on the run) of the form

expected to find inventory collection: TimedOut(30.057651591s)

This is probably also the cause of #8738. It does not appear as often in CI (probably because of the much lower degree of parallelism), but does occasionally, e.g., #8755, #8658. Running locally with reduced parallelism via -j also reduces the failures: with -j4 or -j8, the failures mostly vanish, but they appear again at -j16 and higher.

Here's a sample failure log. It may not actually be very useful for debugging, though, because of #1248.

Reproduce with cargo nextest run --no-fail-fast at varying -j values, using the test_blueprint_edit integration failure as a reference. Compare the behavior around commit #8603 and inspect the timeout logs, while accounting for the debugging limitation noted in #1248. Done means the high-parallelism batch no longer produces these inventory-collection timeouts.