NPN or PNP VAS

The choice is often dictated by the transistors in the front end (differential input stage). D Self uses PNP's for those I believe in order to minimise noise contribution and that means an NPN VAS must be used. Rod use's NPN's for the differential input and so needs a PNP VAS.
 
In the past, good video amplifier/CRT driver transistors were NPN. They made good VAS’s. Still do, but you can get both polarities (C3503/A1381). And lower noise input transistors that could handle any voltage were PNPs. Very low noise NPNs we’re available, and some had gain approaching 1000, but weren’t suitable for very much voltage. The 2N5087 was high enough voltage to be useful and low enough noise. The 5089 couldn’t handle as much voltage, and the 5210 came later. Many of the good Jap input devices were PNP - at least initially - some of those types are on the “replace” list because of worsening popcorn noise over time. None were as bad as the famous NPN 2SC458, though.

Today you can pretty much pick which you want - depending on what polarity of the input bias current works better for you. For singleton input when negative rail is ground, PNP input/NPN VAS still works best for PSRR.
 
The differential pair should use high-gain transistors. In the past, some PNPs have had higher Hfe than their NPN complements.
Ed
A PNP differential has then the advantage of using a NPN VAS, wich at the time were way cheaper and common
than high voltage PNP for this stage, for some reason when there was a PNP complementary it was invariably 2-3x
more expensive than its NPN counterpart.