So I took the time to reload the data I described yesterday, that we had fed into Aurora RDS, into a local MariaDB installation and I also eliminated split/cat from the equation by just using a file that had come straight from a mysql dump via an aws s3 sync operation and the results are disturbing:

Source db (vanilla mysql on our own hardware)
mysql> select count(*) from a_big_table;
+----------+
| count(*) |
+----------+
|  2137447 |
+----------+

MariaDB on our own box
MariaDB [banks_production]> select count(*) from a_big_table;
+----------+
| count(*) |
+----------+
|  2137447 |
+----------+

RDS (from loading same file):
select count(*) from a_big_table;
+----------+
| count(*) |
+----------+
|   797121 |
+----------+
1 row in set (0.00 sec)

That's not encouraging at all.