As i understand it when the SSd cache becomes full it should fall back to the Archive drives
But
If you are copying large volumes of data (internally within the server in this case) it will fail 50% of the time
Basically what appears to be happening is
If the ssd's are beyond their limit % full setting the next file will fall back to the archive drives - as expected - (in my case 95% full)
However if the ssd's are not at or near the % full setting and the next file exceeds the remaining space on the ssd's (large video files) then the copy fails with a STREAM(????) error - not good for unattended large copies 
If you get the pool to balance and create sufficient space the copy will proceed - however this "paused" copy has an unwanted side effect as it generates a large amount of "Unusable for duplication" data in my case over 700GB worth - checking the disks sizes and comparing file size totals - with checking for hidden files and running a full chkdsk /f on all drives did not show any errors or orphaned files - its appears to be a bug in the calc by DP - as copying more data to the pool reduces the "Unusable for duplication" by the amount copied 
. In my case there are no other files on these disks only the pool part dir, recycle bin and SVI - both SVI and recycle bin are turned off and empty as they were the first things i checked.
Have a suspicion that the amount of "Unusable for duplication" is the size of the copy which "fails" and if you have more than one it is cumulative - difficult to check as it depends what files are copied and their sizes to induce the error.
I did reboot to see if it would solve the issue - no joy
I remeasured the pool a couple of times - no joy
i just copied 203GB of data to the pool and the "Unusable for duplication" dropped from 502GB to 299GB - as each file was copied the total dropped by that amount during the copy.
Doing a second copy of 296GB to see if i end up with ~ 3GB left in "Unusable for duplication" - total is dropping by file copied as seen in first copy. [edit] - ended up with 3.34GB unusable left after copy. Now it is back to normal as have copied more data across to the pool and do not have Unusable anymore.
DP version 740 Beta
Win 2012r2 SE
Pool in question is not duplicated
Pool is second pool on server - if that makes any difference
If you need any logs etc let me know 