1. Start Twitmonger with NVDA running and speech output set to "auto."
2. Wait a few seconds to ensure it loads, do ctrl-win-home and up/down arrows, you'll hear the top/bottom of buffer ding but no speech from NVDA.
Same thing happens when SAPI5 is selected in Twitmonger.
Apparently NVDA support works when running from source, though I haven't tested this myself.