EDIT: Additional information provided by LovelyWife(TM) - listening outside the car, she has proclaimed that the rattle / buzz / drone / resonance is coming from the front of the car and sounds like it's coming from below the engine between the wheels.
I was overly optimistic in my assessment of the previous question: How should I diagnose an exhaust "howl" when descending through 3200 rpms?
Here is the current situation on my currently all-stock (intake and exhaust tracts) 2004 Subaru Impreza WRX: when revving carefully around 3250 rpms (+/- 100 rpms), there is a distinct jump in the overall noise level on a totally different peak frequency (higher by about an octave on my uncalibrated ears) from the standard exhaust note. This is a super-annoying drone that is easily sustained and right at a convenient cruising speed.
This isn't present (or is greatly reduced) under acceleration but is very obvious at a steady speed or when descending through that rev band. This induces me to think that the problem is downstream of the turbo as it's most obvious in an off-boost situation.
Here's what I know so far:
- It's not the stock muffler - the problem was much more obvious but still present in the aftermarket muffler.
- It's not the aftermarket muffler - see 1.
- It's not the gasket on the muffler - replaced.
- It's not the nearby heat shielding - same problem with and without the shielding.
- It's very hard to tell if the problem exists with the muffler removed entirely as it's swamped but all the other noise (or song, depending on your preference).
- It doesn't appear to be the donut ring on the intermediate pipe after visual inspection.
- There don't appear to be any obvious holes or leaks in the exhaust tract downstream of the turbo after crawling around under the car all morning.
- Am I hearing the resonator on the intermediate pipe fail?
- Am I hearing the pre-cat on the uppipe fail? This is a fabled bogeyman since the WRX was first introduced to the US.
- If there really is a hidden exhaust leak, why would it be so tightly bound to a single rev range?