Dennis Felsing

Actually when disconnecting SQLsmith the environmentd memory usage goes down again, so it's not a memory leak, but a lot allocated to support the connection.

SLT has lots of failures already, will have to fix/ignore them somehow: https://gist.github.com/def-/45e08b7fd3cd7cc5d88dce6ecc823682

the sum of all messages in the system from [mz_internal.mz](http://mz_internal.mz/)_message_counts_per_worker > The `mz_message_counts` view describes the messages and message batches sent and received over the [dataflow] channels in the system.

Yes, assuming the tests run through successfully.

Another occurrence in CI: https://buildkite.com/materialize/test/builds/84612#019050c7-2b2e-44ac-b591-6e715a3447b8

Seen again in https://buildkite.com/materialize/coverage/builds/236#018ad8d2-4da7-4f15-b52a-66c29d6d8299

Also happens without coverage build, so it's just timing-dependent: https://buildkite.com/materialize/tests/builds/76452#018dd120-3f8a-4bc3-89e6-543915ad4283

For the ingestion lag toxiproxy is good, we already use it in some tests, you can make it as slow as you want in ingesting. Testdrive seems fine for the...

Feature benchmark is showing some memory regressions against main, is that expected? https://buildkite.com/materialize/nightlies/builds/6471

Indeed. I tried bumping the nightly version but that failed because of various reasons, so I tried a few versions inbetween, but they all fail for some reasons. Fixing the...