... and there's a special place in hell for thread hijackers!
Okay, lots of recommendations for hardware and software, but it's too late for hardware and the Audacity software doesn't handle anything quite like this, and I don't expect it's what's usually meant by "noise", so I'm loth to buy software what won't fix the problem.
I have a workable solution, up to a point. It's still very uncomfortable to transcribe from, but my technique at least makes the conversation intelligible. There may be better and easier ways to do this, that don't involve quite so much tweaking though, and if anyone would like to take a crack at a better solution, here's a 30-second chunk to play with:
http://members.shaw.ca/ccfc/Sound_Sample/ - in either .flac or .mp3 format, take your pick. It's a real oddity, the only one with a *double* buzz, but the same technique does work, just not very well.
So, my approach has been to record a piece with the same set-up but no voice, to get just the buzz. I've then inverted that to produce the anti-buzz and matched that millisecond to millisecond to the interview file. You'll find ready-inverted buzz files in the same place. I've used Audacity's envelope tool to match peak heights from one track to the other, in order for one to cancel out the other. In this particular case, with the double buzz, I've added a second envelope-adjusted anti-buzz, and adjusted that to match the peaks of the second buzz. It works - just - , leaving a general background hiss like frying eggs. It needs constant tweaking, though, as every 15 seconds or so the alignment goes astray and has to be corrected. In a 40-minute conversation, that's a LOT of tweaking.
Anyway, if you'd like to give it a try, you're more than welcome. And if you have a better way to do this, I'm all ears!