The 45 was dropped because the output impedance (same as plate impedance) is 2.5 times higher than a 300B. We did a lot of experimenting with different designs to get the source impedance down and the available current up, and have so far concluded that both make an important difference.
As a preamp, a very quiet filament power supply is needed, and it's both difficult and inefficient to regulate 2.5 amps at 2.5 volts. That discouraged the use of 2A3s. 5 volts at 1.2 amps is, in important ways, four times easier. (This is a problem that I will continue to work on ...) Meanwhile, and pending resolution of the last details of the circuit, I think it can easily be made to work with 6A3s, or (with a socket change) 6B4s.
With bigger changes, a few others may be workable - we haven't yet explored the limits of the design.