Unless you have a very unusual border case, CIM-mapping is done in search-time, means the fields are calculated/extracted/aliased dynamically during the search, not while indexing. So even if it's done during the search phase executed by the indexers (as per map-reduce algorithm), it's still done according to props/transforms rules pushed to indexers in knowledge bundle from the search tier.
So long story short - you don' t _need_ to configure your indexers in any particular way for CIM compliance.
However, it's often used practice to define everything regarding a specific sourcetype (including both search time and index time settings) in one add-on and distribute this add-on across your environment - both search tier and indexing tier, possibly an intermediate HF tier if you're using one. This way you have a consistent config over all tiers and only the settings relevant at any particular point in time are used by Splunk. So your index-time settings are quietly sitting on search head tier server and are simply not used because they're not needed. And the same goes for search-time settings on indexer tier.
