In my experience, there is a lot of inaccurate information about DCC. Even the definition of what is considered a keep-alive circuit is not clear. Is 200µF a keep-keep alive? How about 500µf, or 1000µF? Or 5,000µF? 10,000, 200,000µF
I also don't buy that Paragon4 has on-board keep alive much larger than Paragon3? Examining
N scale Paragon3 and 4, Paragon4 does not have significantly greater number of MLCCs (Multi-Layer Ceramic Capacitors) than Paragon3. Why MLCCs? because those are wired in the place where a keep-alive would be connected. These however don't add up to more than few hundred µF (on both Paragon 3 and 4).
In my experience (and also anecdotal evidence gathered on Internet), BLIs decoders in general are difficult to program on programming track. This also depends on the brand of DCC system used.
So to me this while hoopla that N scale Paragon 4 has much better "keep-alive" is just bunch of you-know-what. At least as far as the on-decoder hardware goes. Maybe BLI has changed the decoder's internal programming (firmware) to be more resilient and not reboot so easily? I really don't know for sure.
As for the voltage, the voltage used on the programming track should be 12-14V, which is similar to what is used on main line. It also somewhat depends on the brand of DCC system.
I'm so happy that BLI finally saw the light (and probably bowed to the pressure from the modeling community) and decided to start selling their (pretty darn good models) without their decoders (stealth).
My question is: is the keep-alive and the enable-disable feature mentioned anywhere in the BLI's Paragon4 manuals?
EDIT: Looks like I found it.
http://www.broadway-limited2.com/support/techSteam(0.04).pdfCV221
Unique Control One
Bit 5: 0=keep alive active; 1=keep alive off (HO Only)
So it appears that H0 models with Paragon4 have additional keep-alive circuit built-in. If yours is an N scale model (which I assume since you posted this in the N/Z section) this thing doesn't even apply to your model.