Same Gauss-Kruger zone, two datums, 427 metres: read the proj4text, not the label

The Rajasthan result is worth working out by hand, because it explains why that class of mistake is cheap. Zone 7 carries a false easting of 7,500,000 metres ^[3]; the Slovenian grid carries 500,000 ^[4]. Feed a legitimate zone 7 easting of 7,457,000 into 8686 and you have declared the point 6,957,000 metres east of the 15th meridian ^[1]. PROJ does that arithmetic without complaint and hands back a longitude of 76.6 degrees ^[7]. The absurdity is in the input, and its size is what saves you.

The pair that survives once the label is corrected has no such property. According to the developer's dev.to write-up, 31277 and 3909 agree on central meridian, scale factor, false easting and ellipsoid; the only difference is that one names PROJ's built-in Hermannskogel datum while the other supplies an explicit three-parameter shift to WGS 84 ^[13]. Subtract the two published outputs and the disagreement is 0.005390 degrees of longitude against 0.000022 degrees of latitude ^[2]. At Belgrade's latitude that is about 426 metres east to west and roughly two and a half metres north to south ^[3]. So the 426.73 metre separation ^[14] is very nearly a pure sideways slide, which for a trajectory driven along a road is the least helpful geometry available: the line keeps its shape and its heading, holds a constant offset, and still reads as a survey.

The registry helped set the trap. The Balkans zone codes run 3907, 3908, 3909, 3910 for zones 5 through 8 ^[9], and the later 867x batch reissued zones 5, 6 and 8 but not 7, because 3909 already held it ^[10]. Someone scanning that numeric range for a modern zone 7 code finds a gap exactly where pattern matching expects a hit, surrounded by codes that are real, on the same datum family, and in the same corner of Europe ^[11]. Nothing about 8686 fails a plausibility check. Only the parameters distinguish it, and they say Slovenia ^[5].

Which is the transferable part. The text in a picker is a string someone typed once ^[2]; proj4text is what PROJ will actually execute ^[16]. Those two artefacts are maintained by different processes, and only one of them is under version control by the transformation engine. When they disagree, the map does not argue with you.

Note also that this dropdown cannot simply be collapsed to a single entry. 31277 and 3909 both legitimately describe Gauss-Kruger zone 7 ^[13], and 3909 is the correct modern code for the region ^[12], so a surveyor delivering data in the older Hermannskogel realisation needs the older code. The picker has to ask which datum the file was produced under, not which zone it is in. That is a harder question to put in a form label, and it is the one that carries the 427 metres.