> prefer PNPs rather than the ol' tried and true NPNs
If you remember 1958: all we could get (or afford) was a few PNP types. We made it work.
Then when Sillycon was developed, for a long time NPN was all you could get, or at least NPN were a dime to a dollar cheaper.
There are people who like the old BBC and RCA Ge PNP mike preamps; early Neve Silicon that was all NPN, and people who play with the Flickinger and other early Silicon where they used as few PNP as possible. It was mid-late-1970s before we could get either polarity with good performance at good price.
In many fields, you only need one polarity and it hardly matters which. AM radio. Digital Logic. RTL and DTL were totally NPN jobs; I think TTL is total NPN. With NPN, it is easier to think with negative ground, hence the fad for +5.3V supplies for digital (though there was ECL).
Such fashion can dominate new topologies. Before CMOS there was MOS and a lot of that was PMOS. It was RTL implemented with P-type MOSFETs and wired upside-down on a +5V supply.
Negative ground power supplies have dominated since tube days. NPN often makes the most sense in that world. But anything was possible. When we had radio tubes that would do RF and IF and low-audio with a +12V supply, we made cars with 12V positive-hot batteries, and wired the one Delco PNP Germanium POWER transistor upside-down, collector toward ground. That was easy because the ~5K driver tube needed a transformer to couple to the ~5 ohm transistor input, and just as easy to wire the secondary to +12V where the base and emitter pins were. In fact you could ground the collector to the heatsink and take output from the emitter, though still working the transistor as common-emitter.
99.8% of the time, choice of PNP or NPN is suggested by available power polarity and simple thinking, existing designs and boards, or penny-pinching costs.
It is sometimes a great blessing to have both PNP and NPN available, roughly complementary, at similar prices. (But it is interesting how far you can go with just one polarity, or one good polarity and a low-spec "complement" implemented as a parasitic in the main polarity process.)
When you are reduced to sorting through data-sheets and testing samples to find THE BEST, you are forced to take the polarity of the device you pick. While theory may say PNP should have lower noise, that has not always been true of available devices, and when it is "true" the difference may be lost in production spread. NPN may make Beta easier, but in audio systems you often don't need mega-beta. Given that Ft is rarely over 300mHz anyway, higher beta just means beta starts dropping at a lower frequency. If you "depend" on high beta, you run out of bandwidth sooner.