Savepoints stuck in IN_PROGRESS state in beam 2.27

*The problem:*
After triggering a savepoint, it's status is stuck in "IN_PROGRESS" for ever.

*Context:*
Executed with: beamn-runners-flink-1.10, beam-sdks-java-core:2.27, beam-sdks-java-io-kafka:2.27

Using 
```

KafkaIO.Read to consume messages from Kafka
```

*Observations:*
* checkpoints are no longer being executed.
* files of savepoint are being created on external fs correctly
* no exceptions visiable in jobmanager logs nor in taskmanager
* upgrading flink from 1.10.x to 1.11.3 doesn't resolve the problem

The problem exists after upgrading version of beam components to 2.27.

From flink console:

!image (11).png!

Above image presents the state of after executing savepoint. 
We can see that first operation is being fiinshed properly but ReadData/KafkaIO is still running which ( most-likely) is the cause of why savepoint stays in IN_PROGRESS state.

Imported from Jira [BEAM-11656](https://issues.apache.org/jira/browse/BEAM-11656). Original Jira may contain additional context.
Reported by: wojtekw.