With the ATS clusters ready to serve production traffic (T207048), the next step is to serve a portion of the cache_upload requests via ATS and observe its behavior.
To do that, we need to:
- add a feature flag to point individual Varnish frontends in core DCs to their dc-local ATS clusters
- flip the switch on some cache_upload frontends, both in eqiad and codfw
- observe and document any issues, gather operational experience
- switch back to Varnish