From the scope:

• Expose metrics in serverStatus and log when an out-of-order key is detected in an index build
• Log when a secondary crashes because it’s unable to abort an already committed index build
• Expose metrics in serverStatus (already exists killedDueToInsufficentDiskSpace) and log when an index build has been canceled due to a lack of disk space

Consider if adding an "aborted by secondary" would make sense.

We already log and crash when there's a commit/abort conflict. https://github.com/10gen/mongo/blob/29dec97a718571dbef8ac3ffe4c68115e5874909/src/mongo/db/repl_index_build_state.cpp#L382