## Summary <!-- Briefly summarize the bug --> After upgrading GitLab Runner manager to `18.10.0`, artifact uploads started failing in jobs that run on AWS Fargate because the gitlab-runner binary inside the Fargate task was still on 18.9.0. Runner `18.10.0` now invokes `gitlab-runner artifacts-uploader` with a new `--timeout` flag ([from the work on this issue](https://gitlab.com/gitlab-org/gitlab-runner/-/work_items/39080)), but the `18.9.x` helper binary does not recognize this flag and exits with: ``` flag provided but not defined: -timeout NAME: gitlab-runner artifacts-uploader - create and upload build artifacts (internal) ``` This affects mixed-version environments where the runner manager is upgraded first but helper binaries in job images (Fargate) lag behind and cannot be updated instantly. <!-- What actually happens --> ## Expected behavior The runner manager should avoid passing CLI flags that are not supported by older helper binaries, or otherwise detect helper capabilities and degrade gracefully, so that upgrading the manager does not immediately break jobs where helpers are still on 18.9.x. <!-- What you should see instead -->