Your doubt is completely fine.

Please note that, MongoDB do not undo write operation on the successful written nodes members. It wait until the wtimeout time and return the writeConcern error or successful.

Technically, writeConcern is a acknowledgement mechanism if your data written to the how many member of mongod should confirm receipt to be consider as successful by the driver.

If a write concern fails, it often means that the write was accepted by the primary, but the requested level of acknowledgment was not recieved. Incase if you need only write should happen if writeConcern is successful, unfortunately there is no built-in option to "undo" writes if the write concern fails. Your application logic must check the write concern response and handle potential partial success if strict durability is essential.

Thanks, Darshan