Yes. That is due to method used to generate fluid simulation.
In order to skip computations of too many particles necessary for fluid simulation, particles are created and deleted on the fly.
Amount of particles is readjusted at each frame, from position of Narrow Band at previous frame.
So, if liquid is turbulent, a particle may die when colliding with boundary of domain but several may be generated to replace it in corresponding cell, at following frame.
And as a result, Narrow Band is interpreted as having a too dense amount of particles at boundary, and is expanded to the interior of domain, in order to respect asked density.
So, amount of fluid is increasing, pushing Narrow Band a little bit higher, at next frame.
So, the ratio between cell size (resolution divisions) and particle size is important relatively to allowed amount of particles ( Narrow Band Width + Sampling + Particles Minimum + Particles Maximum).
It is typical of a Particle Radius too big for amount of particles involved. On the opposite, if Particle Radius or amount of particles is too small, you can obtain a disappearing fluid.
So, best way to find correct settings is to debug simulation only looking at particles.
You can improve simulation quality by disabling adaptive timesteps and decreasing CFL Number at cost of a longer baking time.
But here, a simple solution may be to avoid collisions with Domain boundaries and only keep Bottom option enabled in Border Collisions subpanel.