It's the right link and it is Fig 11. Quoting F. Langford Smith, p1283, Radio Designer's Handbook available as free download here;
http://www.tubebooks.org/technical_books_online.htm
"Parasitic oscillations often occur at frequencies of the order of 100MHz with leads from valve sockets forming resonant circuits, and with a feedback path provided by a few pF of capacitance between plate and grid circuits."
Placing a capacitor between control grid and earth, but remote from the socket, won't eliminate parasitics under all operating conditions because there is still a good lead length. The 100R resistors and 270pF capacitors soldered right on each of the valve pins do.
It's normal design practice to use grid stoppers soldered as close as possible to g1 and g2 (and in extreme cases to anode) socket pins but this isn't possible here.
Proof of concept can be tested before committing, by testing, for example, an EF183 or EF184 and soldering just three RC pairs with short leads to g1, g2 and a (pins 2 and 8 and 7) of the B9A socket base. They should have no effect on DC voltages used during leak testing etc. or on your 1KHz test signal but I confess that I haven't fully studied the design. Values aren't critical and cheap 5% 1W resistors and 1000V 220pF ceramic capacitors would suffice.
On my curve tracer design, there were extreme current and voltage surges, or constant current effects on the XY recorder without these in place when testing EF183, though I could successfully trace ECC82 characteristics. My first thoughts were to build an adapter with 240R stopper resistors soldered close to each pin with small switches to short out those not required. A clumsy solution. Then I came across this thread researching other options. I continued, wondering how it was done historically and came across the simple Philips design. Tried proof of concept, and it worked.
Regards,
Bob