Why do you even want a sine-wave?
Most scoping works better with a triangle. You can't see if a sine is bent nearly as easily as you can see a bent triangle. Clipping is instantly obvious.
And a tri-wave is dead-simple to generate. (OK, not as simple as it looks, but simpler than a really-clean sine.)
For frequency response, yes you do need a sine, but it needn't be hyper-low THD. IMHO, those lamp-stabilized things don't work. Yes, I know thousands of audio oscillators with lamps have been sold. The ones I have gotten intimate with would not stabilize unless there were some other nonlinearity beside the lamp. And in that case, why not just hang a Zener on it?
+29dBm???? (+26dBu into 300Ω) That's a power-amp, not a precision signal source. You don't want to generate a Watt and then pad it down. Heath did that; but H-P generates 10V at 10K and then runs a couple 6F6 push-pull to generate +30dBm in 600Ω (and 2 or 3 watts in lower loads). Find an idle but clean 40-Watt speaker amp and put a 30Ω 10W resistor in series with the output: you will get your +26dBu without strain into any line input, and dead-shorts won't hurt it. I actually added a 5-Watt booster to my Heath so I could run speaker tests without dragging out a power amp, but that was in days when there were not decent power amps in every garage and attic.
Is the old 8083 function generator chip still around? The sine was glitched but usable, and the triangle and square waves were there and clean. One-resistor sweep, too.