Description of problem:
QEMU is a large, complex program with many features.  Not all of them are sufficiently useful to customers to justify their support cost. To avoid "you shipped it, you support it" SNAFUs, we must choose the features to ship carefully.

My shot-from-the-hip take on what needs to be done:

1. Identify what exactly to support.

2. Compile out as much of the unsupported stuff as possible.  This may require significant upstream work to improve modularity and compile-time configurability.  Any leftovers will either require non-upstream patches (bad), or "this is not supported" user documentation (worse).

3. Tests to verify the unwanted stuff is gone.

This is almost certainly too much for a single person.  Whoever gets assigned this bug should coordinate a team effort.