This change is 

This test runs 10k times w/ stress w/o flaking now.

Re: can we ensure that all nodes in the cluster see the updated cluster setting value? Perhaps we could adjust the logic test to explicitly read the cluster setting from each node (with nodeidx directive) and with a retry option to wait until the setting is propagated, but we run this tests in 1- and 3-node configs, so we'd also need to skip some of those reads. In short, adding the retries seems like the easiest option.

So basically the read side is a problem, there's a rangefeed on everynode that updates the in memory settings cache. What I'd like to see is a primitive to wait for range feed invocations to be done or something, ie:

# Bounding box operations.
statement ok
SET CLUSTER SETTING sql.spatial.experimental_box2d_comparison_operators.enabled = on

statement ok
SELECT crdb_internal.await_range_feed_progress_on_all_nodes(crdb_internal.get_lastest_timestamp("system.settings"))

But I have no idea how that would work, presumably we'd read the latest timestamp from that table and then check that all the other nodes range feeds have advanced to that timestamp? There's some discussion here:

#87201

Basically I agreed with Yahor retries are the easiest option but this feels like a problem that could come up again and could use a better solution.

bors r+

Build succeeded:

• Bazel Essential CI (Cockroach)