To prevent such mistakes, the user should actually execute the display read data step and the display converted data step. If he had done this, then he had seen that instead of english text there is just a plant number and instead of spanish text is only a vendor number. A little sensitive for data should the person be who has the authority to do mass uploads.
How else would you expect to validate a long text field?
It would certainly be easier in case of using the source for LSMW1 in LSMW2 execution, because here you could just do a select on T001w and LFA1 to chck if the data is valid.
But pure text fields....this can only be decided by thoughtful reading .