I can imagine that it would be very handy to be able to output in tree sequence all the intermediate nodes produced by SLiM - i.e. not just the coalescence points, but all the "unary nodes" too. This would be somewhat intermediate between "saving everybody" and the normal output. Individuals would still not be saved, of course, unless they happen to be "retained" - this is just about saving the node info (genomes) themselves.

The way I imagine this working would be to add an optional keep_unary arg to treeSeqSimplify(), which could be T, F, or NULL (default). There would also be a keep_unary option to initializeTreeSeq() that would default to F and would set the default keep_unary flag during automatic simplification. This value would be the default used when calling treeSeqSimplify() with no params.

What do you think @petrelharp ? I guess you're the one doing "remember everyone" simulations.