I really do not think that a non-linear scale for rail/road/tram vehicles is a good idea: not only would it involve re-scaling nearly every single vehicle of that type in the pakset (and for rail vehicles, there are vast numbers) as the existing vehicles are not scaled in this way, but it would also make small vehicles look very out of place next to the larger vehicles. It would also cause real difficulties with vehicles, such as railway carriages, whose only real distinction from other nearly contemporary railway vehicles is the length (such as the 48ft/54ft LBSCR carriages); the size difference between these carriages will be inadequate if a non-linear scale is used.
There is also the issue not hitherto mentioned that the all the standard gauge railway tracks are scaled in their width to the same scale as the rail vehicles' length (I have not checked the narrow gauge tracks). If rail vehicles are to have a different scale on the x and z axes than on the y axis, should the rails not also have the same scale on the x axis? This would have the advantage of reducing the apparent spacing between the tracks for double track configurations. I should also note that, as things stand, most rail vehicles' wheels are far to far apart for the rails.
As to Vladki's point about noticing poor scaling: it's not an easy thing to keep from people - I certainly notice it, and there is no reason why any discerning user of the pakset might not come to notice it and think less well of the pakset thereafter because of it. It is best to try to prevent this from being a problem in the first place than hope that people do not notice.
As for measurements of early rail vehicles, here is some information on the trams from the "snippet" thread:
P. 79 gives some interesting information on some early (1875) designs of tram engine, these destined for Paris (whose steam tramway was not a success and which reverted to horse operation before electrification in 1896), but designed and built by English company Merryweather & Sons Ltd.. They were vertical boilered engines, although not to the Wilkinson patent, weighing 2t with dimensions of 1.6m long, 2.01m wide and 3.35m high; it had 5" x 9" cylinders with a 90psi boiler. The heating surface and firegrate areas are not given. This was the only locomotive built to these dimensions, the second locomotive, also sent to Paris, had 6" x 9" cylinders and weighed 4t. This latter type of engine became the Merryweather Type 1, and an example belonging to the Wharncliffe National Rifle Association, which used it in an annual makeshift tramway on Wimbledon Common for many years (p. 80).
...it is reported that the (standard gauge) Drypool & Marfleet Steam Tramway Co. paid between £500 and £600 apiece for some Thomas Green tramway engines in May 1889 (a caption on a picture on p. 41 gives a figure of £604 per engine). These were compound engines of two cylinders in an 0-4-0 arrangement with a 5ft wheelbase. Boiler details and cylinder dimensions are not given, but the overall length was given as 11ft 6in, a width of 6ft and a total height of 10ft 4in (not including chimney) (p. 39).
(These are probably equivalent to the Kitson no. 2 engines in the current pakset).
Some information on the size of electric trams by way of example:
In 1905, the LCC ordered a new class E, and built 300 of them (page 41). These were of a bogie type and fully enclosed from new (apart from the driver's position). These were also 33ft 6in long, but only 15ft 9in high. The bodies were apparently of "sturdier construction" than previous cars (implying higher weight), but actual weight is not given. Most class Es were fitted with (again, I assume two) 42hp motors.
Classes F and G were single deck cars seating 36, which ran through the Kingsway tunnel before it was enlarged (p. 41). They were 33ft 6in long, ran on bogies, and were 11ft high. Power and weight are not given.
Edit: Another unverified source, this 'blog comment gives dimensions for the Feltham tram as 40′ 10″ long and 7′ 3″ wide.
Some examples of the dimensions of some early steam locomotives (all page references from "The British Steam Railway Locomotive from 1825 to 1925" by E. L. Ahrons"; where wheelbases are given in two parts separated by a "+" sign, this designates the separate distance between each pair of wheels from front to back):
* based on a scale drawing on p. 36 and my (somewhat crude) measurements of it, a Hawthorn type 2-2-2 engine of about 1837-1841 was just under 17ft long (not including buffers or tender);
* the first "long boiler" engines, which appeared in 1841 on the York and North Midland Railway, had a wheelbase of 6ft 4 1/2in + 4ft 4 1/2in in a 2-2-2 configuration and a boiler barrel 11ft 3in long (this was a 5ft gauge railway);
* a Crampton type engine built for the Eastern Counties Railway in 1848 had a wheelbase of 15ft 3in (2-2-2) (p. 72);
* the LNWR "Cornwall" as originally built in 1847 had a wheelbase of of 12ft 11in, and, as rebuilt in 1858, of 14ft 10in (p. 74);
* the LBSCR "Jenny Lind" type had a wheelbase of 7ft + 6ft 6in and driving wheels of 6ft diameter and carrying wheels of 4ft (p. 77);
* the Metropolitan "A" class of 1864 had a wheelbase of its leading truck alone of 4ft, 8ft 10in of its driving wheels and a total wheelbase of 20ft 9in (p. 155);
* a 2-4-0 LNWR express engine of 1866 had a total wheelbase of 15ft 8in (p. 169);
* the MR 156 class of 1867 had a total wheelbase of 16ft 6in (p. 170);
* the GNR Stirling 8ft Single 4-2-2 of 1870 had a total wheelbase of 22ft 11in (p. 185);
* the Midland 0-4-4 tank engine of 1875 had a total wheelbase of 21ft 9in (p. 201);
* the LBSCR "Terrier" 0-6-0 had a wheelbase of 12ft (p. 202), although note the overhang at the rear; and
* the LBSCR C class 0-6-0 goods engines of 1873 had a wheelbase of 7ft 9in + 7ft 6in with 5ft dia. wheels (p. 203).
I hope that this will suffice for now; as indicated earlier, wheelbases are easier to come by than lengths for older locomotives, but this can be a satisfactorily accurate means of measuring length if dimensions are kept proportional.