Fix flaky test_compute_derivative_rows_mem_usage

[AAraKKe]

What does this PR do?

Replaces the tracemalloc peak-memory threshold check in test_compute_derivative_rows_mem_usage with a direct assertion on the contents of _previous_statements.

Motivation

The test was failing on master (job link) with a measured peak of ~105 MB against a 10 MB threshold, on Python 3.13 with ddtrace profiling active.

The root cause is that the old approach measured tracemalloc peak memory, which is sensitive to:

• Python version (constant folding behaviour changed in 3.13)
• GC timing (whether previous rounds' rows are freed before new ones are allocated)
• ddtrace profiler keeping frame/object references alive longer than expected

Why the new assertion is correct

The property we actually care about is that _previous_statements does not hold on to large non-metric data (e.g. full query strings) between check runs. compute_derivative_rows achieves this by explicitly building the cache from only the metric columns:

self._previous_statements = {
    row_key: {col: row[col] for col in metrics if col in row}
    for row_key, row in merged_rows.items()
}

So after a call with metrics = ['count', 'time'], the cache looks like:

{'sig3767': {'count': 100, 'time': 2005}, 'sig4317': {'count': 200, 'time': 4010}, ...}

The new test directly asserts this: each cached entry's keys are a subset of the requested metrics — meaning the 3000-char query strings, errors, db, user, etc. are never retained. This is deterministic, independent of Python version and profiler state.

Review checklist (to be filled by reviewers)

• Feature or bugfix MUST have appropriate tests (unit, integration, e2e)
• Add the qa/skip-qa label if the PR doesn't need to be tested during QA.
• If you need to backport this PR to another branch, you can add the backport/<branch-name> label to the PR and it will automatically open a backport PR once this one is merged

[AAraKKe]

Added no-changelog since this is only tests. Will involve DBM in the review as well to make sure this is not breaking something I am missing.

[AAraKKe]

This is added to ignore any review we do with Claude in the repo, since worktrees are created it can become a pain to ensure none of this is considered.

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 90.26%. Comparing base (969b7f0) to head (578371b).

Additional details and impacted files

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.
• 📦 JS Bundle Analysis: Save yourself from yourself by tracking and limiting bundle sizes in JS merges.

[eric-weaver]

Lgtm 👍 this does change the goal of this test which was originally introduced here, but given the non-deterministic behavior of tracemalloc on CI this is a better test