Referring to ticket ID #298548, we are requesting Fivetran to improve on the current single Delete operation during sync. 

This single, large DELETE transaction triggered a huge consumption of resources (physical disk space) for two main reasons:

1. Massive Undo Logs: MySQL's InnoDB engine writes a "before image" of all deleted rows to the undo logs to allow for transaction rollback. Deleting 1.9 million records in a single transaction caused these logs to swell to over 2TB. This space is not released until the transaction is fully committed or rolled back, a process which can be extremely slow.

2. Binary Logs: If binary logging is enabled, this single large transaction is also written to the binary logs as one large event, further contributing to disk space consumption.

To prevent this issue, I believe the DELETE statement generated by the Fivetran connector should be broken into smaller, more manageable batches. Instead of a single, massive transaction, using a LIMIT clause to delete a fixed number of rows at a time would prevent the undo logs from growing to a size that overwhelms the disk.

This approach would significantly reduce the memory and disk I/O overhead, making the resync process much more stable and preventing the server from running out of disk space.