In my experience the tree supports are not reliable and break on prints. They also dramatically increase the number of retractions needed to print all those islands and slow down prints. They are ideally suited for SLA which is where they are mostly deployed. (See formlabs slicer for an example)
I print a lot of large parts and throw away a lot of support material. I think a graduated approach would replicate the tree structure in a FDM friendly way. For example, assume 100 layers between the print bed and the first bit of the model that needs to be printed. I setup 3 dense infill layers at 80% infill, 2 base layers, and a 10% default support material. A control like what is currently available for the Temperature settings would be ideal for this with gradual transitions from the various infill levels being applied automatically.
Layer 3 - 10%
Layer 80 - 10%
Layer 90 - 60%
The support would print at 10% and then over layer 80 - 90 it would gradually increase from 10% to 60% giving my dense support layers a nice base to print onto. With this approach very sparse latices could be created without requiring extreme bridges for the dense infill.
+1 for tree supports. I would go one step further and ask for "smart supports". S3D is weak in the calculation and placement of supports. For example, take the case of a recessed hole for a bolt. When the recess originates at the build plate (bottom) there will be a smaller hole in the roof above. The squiggles that are generated are not in consideration of the geometry of the hole, they are squiggles. An "x" would be much better with the center of the "x" aligning with the center of the bolt hole.