I think it is most important that the tts be put in the official repository without the Nvda binary until the issue is settled.
That being said, can't the source code for nvda be gotten and compiled?
It is publicly available.
Nvda is the only speech I use.
It is very stable with no crashing for me using the nvda next snapshot.
The only time the tts defaults to sapi is when I quit nvda.
When I install the tts, my nvda speech works with it just as it does outside of the tts.
As far as braille displays, working with Nvda, many do so very well.
I have used a brailleant with no problem.
I would be happy to help with nvda any time.
Tremendous thanks for everyone's contributions to the tts.
All participants make it possible.