Cluster vs. Index Segment Size Settings Precedence for max_merged_segment

Elastic Stack Elasticsearch
1

I've noticed what appears to be cluster settings taking precedence over index-specific settings for segment size limits, and I'm seeking clarification on how these settings interact.

My index settings explicitly specify:

"merge": {
  "policy": {
    "max_merged_segment": "0b",
    "segments_per_tier": "10.0"
  }
}

While my cluster settings specify:


"merge": {
  "policy": {
    "max_merged_segment": "5gb",
    "max_time_based_merged_segment": "100gb"
  }
}

-I noticed that despite the index setting max_merged_segment=0b (which should mean unlimited), all segments consistently cap at ~5.3GB
Questions

  1. Is it expected that cluster-level max_merged_segment settings take precedence over index-level settings?
  2. Does setting max_merged_segment=0b at the index level not actually mean "unlimited" as I expected?
  3. Why are my segments capping at ~5.3GB when the cluster setting is 5GB? Is there some overhead calculation?
  4. Is this precedence behavior documented somewhere I might have missed?