The problem is currently discussed here:

https://github.com/osmandapp/OsmAnd/issues/22964

Looking at the history, this crossroad was already correctly (from the logical / navigational perspective) mapped 2 times and 2 times broken again.