I am very happy to announce the release of our report (from the project) to the Survey Cadence Optimization Committee (SCOC) addressing a multiple of survey strategy questions.
You can find this report at
https://pstn-051.lsst.io/
The SCOC charge and membership is online at lsst.org – note in particular there is a one-page handout describing the general timeline for cadence recommendations between now and operations.
To try to keep report-particular questions in one place (and to let everyone know why there will be a minor update to the text 
):
Here is a question from @ssmartt:
Hi Lynne and Peter, thanks for posting the report.
Question for you as I was checking the metrics - page 14 on the SRD metrics. Understandably the high level SRD metrics of area, number of visits and errors on parallax/pm are considered. But you also have an SRD metric of “number of rapid revisits (on timescales of between a few to 40 sec) per point on sky”. Is that a typo ? Do you actually mean minutes ?
If it is seconds, then difficult to understand how this plays with snaps and visits (for moving object identification)
Answer:
Oh, whoops -I think you have found a typo. The SRD requirement it is referring to is this:
3.4.0.2 Distribution of visits in time
The LSST will be capable of observing 20,000 deg2 of the sky in two bands every three
nights. While the data obtained with such a cadence will contribute greatly to studies of
optical transients, it is desirable to explore much shorter scales, down to 1 minute. This can
be achieved with frequent revisits to the same field, or by utilizing field overlap. The details of
the revisit time distribution, and its dependence on the covered area, will greatly depend on
the adopted observing strategy and here only rough guidance is provided.
Specification: At least RVA1 square degrees will have multiple observations separated by
nearly uniformly sampled time scales ranging from 40 sec to 30 min (Table 25).
Quantity | Design Spec | Minimum Spec | Stretch Goal |
RVA1 (deg2) | 2,000 | 1,000 | 3,000 |
TABLE 25: The minimum area with fast (40 sec – 30 min) revisits
and basically these are satisfied by field overlaps.
The “uniform” here is an odd part of the SRD definition … obviously, if we have pairs, then it is likely to have a spike at the pair visit distribution time (which falls into this window). We have worked with Zeljko to come up with metric definitions that satisfy the spirit of this requirement (counting the number of revisits less than the expected pair time and more than the expected pair time and looking for minimum numbers of visits in each ‘box’).
And, as @yoachim said,
We don’t usually talk much about the rapid revisit requirement. It typically gets met because neighboring pointings have an overlap area, so as long as we keep most of the slews short, we meet the requirement.
We haven’t experimented much with how dense/sparse the sky tessellation is, mainly because fiddling with it too much would probably make us fail the rapid revisit requirement.
thanks. Understand, so the spirit of the <1min timescales is covered through either the 2x15s snaps (if they are done), and/or the overlap regions. With the ~minutes timescale covered by the visit pairs. And the simulations aren’t driven by these too much.
The simulations aren’t driven by this requirement that much, but the underlying idea (especially the ‘uniform’ part) was intended to influence survey strategy somewhat to get contiguous blocks. In theory, if you did a marching army (raster scan) style series of visits, with a repeat for a pair, you would get a nearly uniform distribution of visit separations. (I think you also have to assume that this raster scan moves slowly across the sky from night to night as well, because otherwise those separations are not in the same spot on the sky).
The snaps and pairs themselves were not intended to be meeting this requirement – it was aimed at overlaps between fields.