The test is bad actually.

https://fiddle.clickhouse.com/9973867d-2aa6-4061-b4c0-8761e5579b84
It exploit a loophole that allows to alter column type Tuple in a way which extends number of elements. Explicit MODIFY COLUMN would throw an error here.

This 2 commands
ALTER TABLE t_rename_alter RENAME COLUMN arr TO arr_tmp;
ALTER TABLE t_rename_alter RENAME COLUMN arr_v2 TO arr;
are converted to 2 alters and 2 mutations
alters:
DROP COLUMN arr
ADD COLUMN arr Array(Tuple(DateTime, UInt64, String, String, String)) (5 elements)
mutations:
DROP COLUMN arr
MOVE arr_v2 to arr (5 elements)

Alter and mutations could be delayed. They are not synchonized.

When we read from the table we could see metadata with column type arr as Array(Tuple(DateTime, UInt64, String, String, String)) but have the files on the disk where column arr as Array(Tuple(DateTime, UInt64, String, String)).
It tries to apply CAST, but it throws the error.

I have added a synchronization. Now all alters and mutations are done before the reading.
When we apply mutations we apply DROP and MOVE to the files therefore we do not threat old column arr with 4 elements in tuple as a new column with 5 elements.

Changelog category (leave one):

• Not for changelog (changelog entry is not required)

Changelog entry (a user-readable short description of the changes that goes into CHANGELOG.md):

...

Documentation entry for user-facing changes

• Documentation is written (mandatory for new features)