Temporarily disable the IFR flag featureFlagChangeStreamReaderV2 to avoid a potential post-batch resume token regression issue in sharded cluster change streams.
When resuming a change stream in a sharded cluster on mongos using a change event resume token (via resumeAfter / startAfter), the change stream was internally resumed with a high watermark (HWM) token instead of the actual event token. The HWM token has the same cluster time value as the original event resume token, but sorts before it. If the stream then returned an empty first batch, the HWM token was returned as the stream's postBatchResumeToken value and not the original event token. This made the resumed stream unsafe to resume from the returned postBatchResumeToken.

The issue happened in the precise shard targeting for sharded cluster change streams and not in the traditional shard targeting.
Disabling the IFR flag makes change streams use the traditional shard targeting and prevents the issue from happening.

The next 9.0 releases will receive a fix for the issue and should have the IFR flag enabled again for precise shard targeting.