This process allows indexes to be built on the primary and secondaries of a replica set at different times, and indeed recommends building them on secondaries first:

https://docs.mongodb.com/manual/tutorial/build-indexes-on-replica-sets/#build-indexes-on-all-secondaries

From 2017 APAC Offsite, it is anecdotally quite common to build indexes only on secondaries, e.g. for analytics workloads.

This screenshot is a test with the replica set connection in the top right Compass window, a direct connection to the primary in the bottom left window and a direct connection to the secondary in the bottom left window:

My intuition is that Compass should show some kind of merge of the two states below, e.g. perhaps a warning triangle next to the index definition to indicate which nodes have the index present.