Early May I reported a bug in a GPOS for Devanagari with additional languages in the script table. At the time (6 May) I send the file to the info account with description. I upgraded from 1230 to 1245, and the bug is still there.
There are anchors and kern pairs in the source. And there is an abvm and a blwm feature for mark-attachment, a kern feature for some of the kerning and a dist feature for most of the kerning. And there is a default deva and default dev2 script entry which get the abvm, blwm, dist and kern feature . And then there are MAR, NEP and SAN languages for deva and dev2 that get the abvm, blwm and kern feature, but NOT the dist feature.
In the mean time the source file has been further developed. Among other things we added a context lookup for the GPOS. Which now has the effect that the default for deva and dev2 get a dist feature with the context lookup and kerning. But for the language entries the dist feature shows up without kerning.
So that learns us that it is not the case that Glyphs.app doesn’t add dist features to languages, it does’t want to add kerning.