How about a new Como roundhouse image circa 1899-1900? Notice the front of the 6-8 stalls from the right are in line with each other and don't follow a set radius from the turntable suggesting salvaged engine house material.
I noticed in Rick Steele's posted D&LG roster of 1893 that Como is listed (page 12) as having 14 stalls: 6 stone and 8 frame, which would appear to be yet another configuration.
I've always been a bit surprised by the orientation of the original 6 stone stalls, as none of the stalls align with any of the approach tracks, but the whole structure faces pretty much directly east. I'm going to hazard a guess that that was an attempt to minimize snow drifts against the doors and turntable pit, but would be curious to learn of any other less conjecture-based rationale.(Edit--deleted comment about the diagram above as a half-circle, since--ooops-- I misread it.) It does look like the average track separation was 10 degrees.
The Model Masterpieces kit I have for the stone roundhouse specifies 12 degrees per stall, which is a good argument for some selective compression when I get to building the frame additions to model Como circa 1910.
To my knowledge the Roundhouse was extended 3 times to reach its 19 bays, which it would have been in 1910
If you look along Park Gulch you can see that at least the third extension must have been built on fill material. So why did they position the first section the way they did knowing the issues that would arise if it was extended significantly.
In crude terms the stone section would need to be rotated about 6 bays clockwise to block the wind blown winter snow. They did have a snow fence incorporated into the 3 bay extension and after that burned in 1935 they did build one across, last week the turntable would have been inoperable due to snow in the pit.
I have never been able to find out why they put it where they did, a long way from a water source, expansion would have been difficult, they had plenty of other options.
I can think of other situations where the head office did not take much notice of local knowledge so perhaps it looked good on paper?
Thanks for the additional info! While it's great to know, I'm thinking about what would be involved in indexing a model turntable for variable separation tracks, and am probably going to fake it and go with uniform separation.
Unless I can figure out a way to make the 'Armstrong' method compatible with DCC.