That pretty well confirms that 100 KiB is blocksize in use, but the table I had in mind was:
That’s maximum size. Ends of files, small files, metadata, etc. are usually less than the max.
Without explicit blocksize, that would be the 100 KiB default making lots of blocks to track.
That’s the default for new backups, but unfortunately old backups must keep old size AFAIK.
Blocks are “known” by a hash of the data, then a file is defined by list of those block hashes.
Theoretically (I think) Duplicati could make larger blocks by looking over blocklist for patterns.
It would then have to get small blocks from destination, concatenate into big block, and hash.
Why the heck CAN’T we change the blocksize? is a long and technical discussion of problem. Current developer was probably unavailable then. Not sure if any easy ways out of this exist…