There's a problem with dbfiddle.uk with the MySQL instances (see the thread [here](https://dba.stackexchange.com/questions/269014/does-mysql-8-ascii-vs-utf8mb4-0900-ai-ci-size-differ-when-only-using-ascii-chara) on dba.stackexchange).
It appears that the MySQL instances give erroneous results for byte and charater count for strings in Asian languages using the utf8mb4 CHARACTER SET - see my comments and Solomon Rutsky's replies in the thread.
The SQL Server fiddle **does** behave correctly!
Rick James has added a thorough post that explains what is happening behind the scenes.
I found the problem. As I had suspected, the following system session variables are all set to `latin1`:
You can see this by executing the following:
SHOW SESSION VARIABLES LIKE 'character\_set\_%';
This is discussed in the MySQL documentation for [Connection Character Sets and Collations](https://dev.mysql.com/doc/refman/5.6/en/charset-connection.html#charset-connection-client-configuration)
This can be fixed either per each session / connection using `SET NAMES`:
SET NAMES 'utf8mb4';
or, better yet, in the config file: