We don't know, but I think to much is being read into that throw...

We don't know, but I think to much is being read into that throw away line. It was in the context of him noting that the retention can be tuned for various requirements and he quoted some different volatile memory options that would give them a performance advantage over current generation use cases where the retention is fixed. The exact maximum achievable retention is something they have kept hidden from as long as I can remember as it was considered highly market sensitive. The impression I have got from how they have described it each time was that it was always sufficient for the test target model they were targeting at the time - i.e exceeded what they needed, but that they had not established it. My take away has been that maxing that will be at the price of speed, but speed was what they were after so they didn't want to sacrifice speed for persistence at that time, but given how he was talking my reading is that they are confident they can do up to a few hours, maybe a day, without a noticeable impact on speed, but higher than that they don't know yet. There is a lot of assumption and reading between the lines in that interpretation but that is how I am interpreting the confusing sidestepping going on about this issue.

It is going to become an issue now, so they are going to have to address it.