Sending 1 BTC back and forth a million times creates a single transaction chain, sending a million transactions of 0.000001 BTC makes a million nearly independant transactions which all must be remembered. Due to the way bitcoin can drop old deeply confirmed transactions the first is far less overhead than the second in the long run. There may be similar network cost but the disk space cost can be greatly reduced for the single chain.
Only if the "dust" is combined back together and confirmed deeply enough again only then can the dust space be dropped.
Does sending 0.1 BTC a million times also create a single transaction? Where is the line when it goes from a single transaction to independent transactions?
Can the line be moved down slightly? Any bad side-effects if it was?
I don't think you understood, it's nothing to do with the size of the transaction. it's the way transactions get split up. If you start with 1 BTC and send 100 x 0.01 transactions to your friend thats 100 transaction chains one step long. He may or may not combine them back together into a single 1BTC transaction but untill he does the net must keep all 100 transactions stored.
If you start with the same 1BTC or 0.01BTC or any single amount and pass it back and forth to your friend 100 times. Its a similar amount of traffic and storage at first but once the last one gets deeply confirmed (ie 6 blocks/1 hr later) you can forget/delete from your disk database all the transactions but the last one. It's the Merkle hash tree pruning stuff in the white paper.