Support

File exclusion by size, while possible, is very dangerous. It's easy to implement as a custom Stack Filter placed in administrator/components/com_akeeba/akeeba/filters/stack. You need an INI and a PHP file, the INI having the configuration interface and the PHP file the implementation. You could clone the dateconditional filter already there, modifying lines 49-62 to check for the size instead of the date. Please note that stack filters make the backup really slow. They are called once for each backed up file, i.e. about 5000 times on a small site.

Regarding Kickstart, since you are talking about JPS archives that would be impossible anyway :) JPS contents are totally encrypted. You need to decrypt the file header block to figure out which file it is, then you have to keep on decrypting file blocks to see where the data ends. Since you have a problem with the decryption taking too much memory I guess such a feature would simply drill a spectacularly big hole in the water :D

Alternatively, you can extract the JPS archive locally using Akeeba eXtract Wizard. You can then compress the files as ZIP using Finder (Mac OS X), Windows Explorer (Win XP and later) or Nautilus/Dolphin (Linux). The resulting ZIP file can be extracted by Akeeba Kickstart just fine.

Akeeba UNiTE doesn't need an S3 import. Being a CLI tool you can use it inside a bash script. Have you heard of the s3cmd command-line utility? It's the cat's meow for managing S3 from the CLI! I have my Raspberry Pi fetch the latest backup of our site every night from S3 using s3cmd, then let UNiTE restore it on a test site. It's working really great.