[Bug]

### Search before asking

- [X] I searched in the [issues](https://github.com/apache/paimon/issues) and found nothing similar.

### Paimon version

0.8

### Compute Engine

flink1.17.1

### Minimal reproduce step

CREATE TABLE test (
`id` BIGINT ,
`user_id` BIGINT ,
`user_source` BIGINT ,
`identity` BIGINT ,
`friend_id` BIGINT ,
`friend_source` BIGINT ,
`platform` BIGINT ,
`delete_type` BIGINT,
`add_time` STRING ,
`ums_ts_` BIGINT )
USING paimon
COMMENT 'test'
TBLPROPERTIES(
'EXTERNAL' = 'TRUE',
'bucket' = '600',
'bucketing_version' = '2',
'changelog-producer' = 'input',
'consumer.expiration-time' = '30m',
'file.format' = 'parquet',
'metastore.partitioned-table' = 'true',
'path' = 'hdfs:///tmp/paimon/test',
'primary-key' = 'id,user_id,friend_id',
'sequence.field' = 'ums_ts_',
'storage_handler' = 'org.apache.paimon.hive.PaimonStorageHandler',
'write-only' = 'true',
'write.merge-schema.explicit-cast' = 'true');

Our PK table is sharded, so we specified the primary key with three fields: 'primary-key' = 'id,user_id,friend_id'. After running for a while, the compaction task reports an error: Caused by: java.io.IOException: java.lang.IllegalStateException: SortedRun is not sorted and may contain overlapping key intervals. This is a bug.

We found that some data has duplicate id values, even though we specified three fields as the primary key. Additionally, we discovered that _KEY_id is NULL, although _KEY_id should theoretically have the same value as the id field. In what situations would _KEY_id be NULL? I believe the above error is caused by _KEY_id being NULL, but I don't understand the reasons or scenarios that lead to this. How should I troubleshoot this issue?



### What doesn't meet your expectations?

How should I troubleshoot this issue?

### Anything else?

_No response_

### Are you willing to submit a PR?

- [ ] I'm willing to submit a PR!

No source files or tests are named. Start by reproducing the provided CREATE TABLE case on Paimon 0.8 with Flink 1.17.1, then inspect the compaction failure and compare the declared three-field primary key with rows where _KEY_id is NULL. Done means identifying the cause or scenario and documenting a concrete troubleshooting path.

Written by the indexing model from the issue text.