false alarm: [ERROR] mariadbd: Incorrect information in file:

During

ALTER TABLE t DROP PARTITION p;

   ALTER TABLE t ADD PARTITION (PARTITION p VALUES ...);

Sometimes, if the table is still open ? from another thread, it will be Observe error log on each node.

*Actual result* On one node the error log contains:```
[ERROR] mariadbd: Incorrect information in file: '<datadir>/<db>/<tbl>#P#<part>.frm'
```

(internally ER_NOT_FORM_FILE, raised from open_binary_frm() / partition open path in sql/table.cc).
The statement itself succeeds, replication continues, the table works normally.
Expected result No [ERROR] should be logged when the operation succeeds on all nodes and the table is consistent.

The only difference on the affected node is the table_def_version value embedded in the FRM, which is a per-node, monotonically assigned ID (not cluster-synced).