R9 allows the tiny bias current of U2A inverting input to flow, a DC
path, so OpAmp output stays in linear region, eg acts as a follower at DC
so the OpAmp output does not saturate into a rail depending on
sign of bias current. And as crutschow points out is part of the
compensation loop along with C4.
R5 can be analytically determined, or in practice most folks pick a
value in the ~ 50 - 200 ohm region. Then look at transient response
as it decouples the MOSFET Cgate and miller from severe C loading
of U2A output which produces inadequate phase margin, ringing, even
oscillation. So it too is part of the compensation loop in a sense. Interesting
is datasheet has this chart which helps you (along with phase response
A useful translation for determining R5 is to translate its series R & C to
an equivalent parallel R & C which allows you to see how much C
is reflected to the OpAmp output. Then refer to C loading graphs in
datasheet to see if you have adequate phase margin.
Summary of the series-parallel transformation for impedance matching in radio circuits