I pushed a change -- via CDS -- to enable track_cluster_stats/request_response_sizes (#11559) and the change went through (e.g.: /config_dump shows that clusters picked up the knob) but still no stats:

$ curl -s 0:9901/stats  | grep upstream_rs_headers_size
cluster.foo.upstream_rs_headers_size: No recorded values
cluster.bar.upstream_rs_headers_size: No recorded values
...

I tried a binary (hot) restart and also nothing... Anyone using this feature? Digging through the code now to see if
there's anything special about it. It's not working for both mTLS upstreams and over the loopback upstreams (not that it should make a difference).

cc: @sc0ttbeardsley