Could they look to add a checksum verification on Amazon end, this would save everyone a pile of bandwidth, but add a pile of compute time on Amazon end I guess.
1. Compute chunk checksum
2. Upload chunk
3. Run Amazon checksum API
4. Receive Amazon generate checksum (will be only a few bytes depending on MD5 or SHA1 etc)
5 compare checksums from step 1 & 4. If same move to next file, if cehcksum is different, error restart at step 1 for current file