I think that third option is my primary one. For small dependencies I'll often use code by copying the relevant source into my project with links back to the original source. It forces me to understand the code base better. Usually I find I can fix the issue. _Sometimes_ i contribute it back.
Contributing code back is tricky for a lot of the same reasons that leaving your open source repo's issue track open is...fwiw I don't use code with licenses that require contributing back.
