The go toolchain has a built-in race condition detector, #6760 introduces it as a non-binding step in the PR workflow. Its output is only logged but errors are ignored because it reports a good number of issues discovered throughout the KEDA codebase. Ideally, these get gradually addressed one by one and then the PR workflow can require race detector to pass before a PR is merged.

List of issues/PRs that relate to reported race conditions:

• fix: ScalerCache gets the lock before operate the scalers #6739
• Fix race conditions in tests external scaler and predictkube scaler #6764