Open
Description
Similar to #78812
/_cat/shards requests the full routing table for all indices it targets. This might become quite expensive for larger states and appears like an unnecessary round-trip as well.
We should be able to respond to these requests without running this expensive request to master. The indices stats request that follows the getting of the cluster state contains all the shard-routing information we require for building the table anyway (except for the shards that failed the stats request it seems, but those could be addressed by taking the state from the local cluster service that we base the stats request on anyway).
relates #77466