fun fact: i dunno what topology is used in 3sis, but it happens that the 2-pole state variable filter does indeed produce a perfect sinewave as its impulse response (with Q=inf.) in fact it produces a sine and a cosine.
in that configuration, the difference equation of the discretized version would be:
// init:
x = 0
y = amp
// update sample:
x = x + c * y
y = y - c * x
where c is the frequency coefficient 2pi * freq / SR, and x and y produce cos/sin.
as far as i know, the main reason this isn’t more widely used as a low-cost oscillator, is that the amplitude of the sinewave varies nonlinearly with frequency (low freq = higher amp.) its a great trick for quadrature LFOs though.
(of course its true, in analog world nothing is perfect, including impulse functions, thank goodness)