Pressurized fluid flow (again) using 'accelerate'

Normally, we’re not very interested in fluid flow, but would like to check water permeability of nanoscale holes in atomically thin membranes (think MoS2, for example). Out of the box, Gromacs provides this through ‘accelerate’ by applying artificial gravity, for which I calculate the desired ‘g’ as P/(rho*h), where P (of order 10 bar) is the desired pressure, rho and h (of order 10 nm) are the bulk water density and box height, respectively.

Here is what happens… Instead of unidirectional flow, water flux goes up and down, which to me looks like hydraulic shockwaves unquenched by the velocity-scaling thermostat during ~500-ns-long simulations. This makes some sense, given that pressure is applied essentially abruptly.
So far, I tried to increase pressure in 50-ns-long steps and also attempted the same thing by pulling solvent with constant force – similar flux oscillations occur.

Question 1: Do these pressure levels even seem reasonable?
Question 2: if the acceleration is entered into mdp as a positive number, is the resulting force / ideal steady-state flow supposed to be positive (toward +z) or negative?
All comments are welcome.