Asking for patterns and precedents to emulate is going to be quite hit and miss, because there is no universal pattern for zooming. There are many different ways of implementing this sort of behaviour, each with advantages and disadvantages for different use-cases. Without knowing what sorts of information you present, how your users expect to navigate it and how they intend to consume it, it's impossible to tell you what interface to use.
That being said, some questions to ask yourself are:
- Does a user, looking at a node, need to know where they are within the 'overall' structure? If so, they might need to see both the 'zoomed' and 'overall' view simultaneously.
- Does a user, looking at a node, need to know how far deep within a structure they are? You can signify level and hierarchy with font-weight and other visual cues.
- What sort of data are users looking for, and how are they going to consume it? Depending on your information architecture, that might suggest that they need to quickly navigate in certain ways, or see some data outside their current node. That will inform your interface too.
- Are users navigating in such a way that they need to move not just between elements and parents / children, but elements and 'cousins'? That will affect the UI too.
I wish I could be more helpful, but unfortunately, pattern matching will only get a design so far. You need to think about what users are looking for and how they expect to look for it. Once you nail those questions, the rest of your UI decisions should become fairly straightforward.