each address/cell in Flash RAM has a limited number of write cycles before it fails. It is not an explicit number but rather, a probability that failure will occure on each write. Statisticly, it may be 100,000 cycles on average, but inactually it could be the first, or it could be the millionth. There is a real difference in quality flash ram and poor quality and it is not how fast it operates (well, that is a signifigent difference, but not what we are talking about here), but rather how many write cycles statisticly, on average, it can take before failure.
Fragmentation is an indication of individual cells that get far more write-cycles than an average location. Writing files is not a uniform process, that all locations share equally. Some files, will stay put and those cells involved will get very few writes. Other addresses, near the end of the address space, will stay empty virtually all of the time and will also get very few write cycles. Then there are the locations where files are constantly being created, and deleted: These are the locations that are prone most to fragmentation. These locations because of ther high frequency of write cycles are the ones that will also be prone to early failure. I would argue, that the only location with a similar chance of failure would be the FAT table because the pointers to the sequential locations of a file changes just as frequently as the files themselves are created/deleted.
Note, what I'm not saying: Defragging the drive will not help the situation: It will make it worse because it just adds to the number of write cycles.
Cluster size will matter because a larger cluster will have more writes when written to. When you write a 64k cluster for a 1k file, then an 63k extra addresses (as compared to a 1K cluster size) are written with zero's and each one of those writes has a statistical chance of failure every time it is written to.
