If you have a list in a nanoflow, you pass it to a sub for some dedicated processing and return an extra validation message, then subsequently the changes made to the list are out of scope. This is problematic, as this makes nanoflows behave quite differently from 1) Microflows and 2) JavaScript.
Why we need this: if we can only return 1 parameter, then sometimes it will be essential to keep certain lists in view for specific editing. E.g. Image a long list of questions, validating each questions, but then when you have a dependent entity outside of the list, the list may have to be expanded. This has broken down in Nanoflows from Mendix 10.12.3, and it should be fixed, in my view ;-)