I don't think there's been an official statement about chained pooling being built into Darksend. AFAIK it's just an idea that was tossed around. Personally I think it's necessary for any usable level of anonymity (preferably with the ability to specify whitelisted masternodes).
I think it was mentioned, in the discussion between evan/anonymint that by darksending it multiple times through a number of nodes it reduces the bad-actor probabilities to a statistically safe degree to solve the issue that the darksend node knows who sends what.
ok -- but with a timing analysis, the number of transactions in (let's say) a 60 second window will be relatively few, so it's not too hard to reconstruct what really happened, especially with uncommon amounts (e.g. I send you 16.023957).
Unless you force people to use 10/1/0.1/0.01 and everything is grouped in batches of that kind and it is stretched out over more time. I think this will be a case of continuous upgrade of the anonymity capabilities (DarkSend 1, 2, 3, 5, 10, 50 etc) with extra features / capabilities / improvements as we go along...