Ah... got it. Thanks for the amplification.
So, to be sure, the data for the database isn't actually materialized at the cloned instance. Does that make these cloned databases "read only" or can they actually be Inserted/Updated/Deleted and only those changes are materialized and only on the cloned instance? Is the latter a correct understanding on my part?
Also, since you said the clone would occupy <40MB, where is the data in the database read from when data is read? Obviously I'm concerned about the possibility of a "bottle neck" and I'm also concerned with the unplanned materialization of data on the cloned instance.
I'm also concerned about "backups" of development code. Having such a thing has saved our hinnies more than once (we treat backups on Dev boxes as seriously as we do prod).
And, to say the least, your good article has piqued more than just a bit of interest on my part.