Similar to SERVER-132715, collMod can get stuck behind the long running index build while holding DDL lock. In this case, when collMod is blocked due to a stuck rolling index build, it creates continuous replication lag.

This creates a challenge as there are no means for the collMod interruption and recovery without performing a node replacement which can be time consuming with a large data set. Please see the issue in HELP-98422 for details if needed.

Can we consider that collMod should be smart enough to detect the index build and abort after a reasonable timeout instead of retrying indefinitely?