Two Ts will work, but really, what you probably want is either a 50 ohm resistor in series with each port, or a 2 way divider. That way, the impedances presented to the UUT will be 50 ohms (not 25 ohms). The 25 ohms would calibrate out, but some UUTs (amplifiers) may not like the mismatch.
I'm experimenting with this right now.
In this case we know "works" was defined by:
No reason it won't work - at least to the performance of the NanoVNA.
With nothing more than Ts to couple the two VNAs, even with the added stubs, poor termination, 25 ohms, running the VNAs asynch, the system will perform to the same level as using one VNA properly.
There was an paper I came across that was talking about common problems beginners have with VNAs. One of the most common problems was the beginner feels they can calibrate out every error.
I remember thinking the person that wrote that paper had some pretty good stories to tell.
As far as
what you probably want
, I went over that some time ago and wasn't able to come up with anything worth documenting. This is why an actual transfer relay was used. Sprinkling in a few 50 ohm resistors isn't something I would consider.
The DPDT may be acceptable at low frequencies with the added unterminated stubs. Even with the cables bent to their tightest radius, that stub makes zero sense to me and I see no reason to take the hit in performance. It may be perfectly fine for the person who posted it.