The DocValuesConsumerUtil class contains static methods that determine whether or not a lucene merge can take an optimized path. Codecs that result in UNSUPPORTED merge stats being returned from these methods will pay a hefty penalty in their merge performance. However, it is tricky to detect this outside of benchmarks, as we don't have a good way of testing them directly - the method calls are deep inside the merge machinery and require a lot of set-up.

We should build some test classes that allow us to easily test if a custom Codec is using the optimized merge path. This would allow us to detect regressions in CI, rather than having to spot changes in nightly benchmarks after the fact.