Indices.breaker.total.use_real_memory or search.check_buckets_step_size  didn't work

I always thought that with this new parameter, requests for too many buckets should be rejected。But I tested it today, making two requests and creating hundreds of thousands of buckets per request。In the end, the machine is still directly OOM。
My ES version is 7.10.1。I tested not only the case where indices.breaker.total.use_real_memory is false, but also the case where indices.breaker.total.use_real_memory is true。And the max_buckets is the default value of 65535.The result is the same。



But when I set the max_buckets to extremely large (instead of the default value of 65535). The circuit breaker will come into play and the cluster will not crash and will not OOM.
I feel very confused. I hope you can help me. Thank you very much