This is nonsense. TF is a predecessor standard name, you’d be unlikely to find to find one remotely large enough. That modern pis use a micro form factor SD card rather than a full size one is not practically very relevant, since the internals are often the same and more than a few first generation pis were run with micro cards in passive adapters.
It is true that an SD card is not a very robust solution for a field deployed embedded system - eMMC or well engineering explicitly managed NAND is better than the near unknowable mystery of what goes on inside a consumer SD card. More compact Linux platforms of the sort chosen for purpose built gateway products can often be run by initializing a ramdisk from robust NOR flash, but it just plain takes too much code to make a pi work for that to be practical. The pi compute modules do have a more useful range of storage options.
None of this likely has anything to do with the actual issue of this thread, which is probably related to how the card image was downloaded and written to the card. In particular some writing schemes want to start with a card containing a single FAT partition rather than one that was previously partitioned for use in a pi.