We probably want to split this in a couple sub-bugs for each wrapper. Advantages: - for a simple wrapper (SOW) this traces 10 lines for hundreds of lines of existing code - non-native wrappers are faster to operate on, no need to locally "shadow" each operation and update the property tree - non-native wrappers use less memory since we don't cache properties locally - we can make expandos survive explicitly where needed, but only when we have expandos do we need storage for them - be more transparent and less leaky - if we start tracing proxies, we will automatically also optimize wrappers

Not to mention, this also eliminates a whole bunch of open bugs we have for wrappers that result from their non-nativeness, i.e. keys() only shows properties that have been resolved so far.