Looks like I need to do structured testing.
From memory, obsidian floor produces black sand soil.
Since I can (and have been in the past) wrong in my recollection of this issue--
I will create a world specifically for this purpose, will save it in its virgin state, at first embark, just before experimental test runs (all known possible ways to produce a soil tile), and just after.
I will package them all into a zip for dffd. This will take some time.
DFhack will be used for expediency to generate mud using water creation.
Experiment log, Site selection.
A world was generated using default parameters. The following site was selected due to the basic conditions of having half the embark in a known "No soil" biome (mountain), and half in a "has soil" biome (shrubland)
Experiment Log, Initial experimental setup
Several rooms have been carved for the experiment. Two in the shrubland biome, 1z below surface level. Extant soil. Two in the mountain biome. No soil.
In each of the target environments, one room is native to that Z level, with native floor type. The other has been channeled one z down, and a floating constructed floor has been produced. A quarry was constructed to provide building materials. Fastdwarf was used to expedite the process.
Same setup, one z below.
No water has been produced at this juncture, so that this apparatus can be saved for further revisions on the test theme.
Shrubland biome -2z lacks soil. This initial test setup can be adapted in later run to drop soil onto layer below to demonstrate soil creation through layer drop effect.
Experiment 1, phase 2
Water is dropped, and mud is created.
Experiment 1 Phase 3
Paved roads constructed over test areas.
Test apparatus saved for fork in testing methodology.
Experiment 1 phase 4 (Fork 1, deconstruction)
Null result-- No soil generated in any zone.
Experiment 1 phase 4 (fork 2, Floor drop 1)
Null result-- No soil generated in target zones.