The issue of thermal coefficient of expansion was recognized by Henry Ford a hundred years ago (Moving Forward, 1930). "Arguments between a mechanic on the warm side of a shop with one on the cold side often became legal battles between seller and buyer which involved broken contracts, months or even years of litigation, and often not inconsiderable damage awards by courts. " The two mechanics, presumably one representing the supplier and the other the customer, were of course telling the truth according to what their instruments were telling them, but the parts on the warm side of the shop were slightly larger than when they were on the cold side.

Ford noted yet another problem; the gauge could itself change over time. "When a piece of steel is hardened tremendous internal strains are set up within the surfaces of the gauge which cool and set first as the gauge is quenched after heat treating. These strains are gradually relieved by slow expansion of the metal—a process that may continue for a year or more after the gauge is completed. Also if a gauge is not used at the same temperature as that at which it was finished its size will vary with the difference of temperature." Johansson gauge blocks, however, apparently surmounted this issue: " Selection of the steel best suited to the manufacture of gauge blocks and the discovery of the heat treatment which would render this steel sufficiently stable required experiments over a period of nine years. Then a universal gauging system passed from an ideal to an accomplishment."

This article discusses technology that was not available in 1930, namely the ability of computers to compensate for thermal expansions of both the part and the gauge when measurements cannot be performed in a climate-controlled lab. Even if a climate-controlled lab is available, it doesn't help when "Workpieces that have just come out of a dry machining process can be several degrees warmer and remain so for hours." If we have to wait several hours for them to cool down, that's wasted cycle time (waiting). Ford got around similar issues by sending specimens to a lab (for metallurgical tests as I recall) while a lot was being processed, i.e. the work did not wait but would be stopped only if the specimens were out of specification, but this doesn't work when all the parts have to be measured before they can go to the next operation.  If temperature compensation can be done at the point of inspection, though, this eliminates the problem completely.