But the program used to test the cards here was h2test which is a USB Flash drive test program for Windows. i.e. it is accessing the device using files.
On Windows, ever SDcard reader I've seen has an embedded device controller that talks via USB and they appear as USB Mass Storage devices using the USB class driver. The SDcard unit has a micro and SDcard host controller and it's translates the USB mass storage file commands into SDcard block read/writes etc. That's not code running on the host per se but the access is via a filesystem access from the host with embedded driver in the card reader. The SDcard knows nothing of this.
If you stream writes to the card to the maximum size of the card you will fill up the space. Reading back and comparing will let you see if the card capacity is what you expect and that you wrote what you intended. You can also seek about the file randomly to check that the card gives you the right data. As long as you only write once, then whether the card does wear-levelling or not wont matter it wont get the chance.
If you want to do direct access to any block on the card you need to access the SD host controller directly and you can't do that on Windows USB SDcard units. You can on embedded things like mobile phones or a Pi where the host controller is accessible from the CPU.