Hi Zalz,
Your fluid properties look fine!
If you're using the latest verison of Unity you can look at https://docs.unity3d.com/ScriptReference/Rigidbody.GetAccumulatedForce.html to find the total force acting on a body - you can then compensate for large forces by adding forces in the opposite direction of the accumulated force. Note you would need a script running in a larger execution order to ensure all the forces have been added to the rigid body before saturating them. Check out this for more on execution orders: https://docs.unity3d.com/Manual/class-MonoManager.html
For your test, you can check the expected terminal velocity of an object using the following equation:
v = sqrt( 2mg / pACd )
Where v is the terminal velocity, m is mass, g is acceleration due to gravity, rho (p) is the density of the fluid (1.23 by default), A is the area (0.785 m^2 for your cube - note we use ellipsoids to approximate bodies) and Cd is the drag coefficient (0.5 is a good approximation for a cube). I get 6.37 m/s for a 1x1x1 m cube at 1 kg. Our drag model is a bit more involved than that equation which will explain the discrepancy, the cube actually reaches 6.33 m/s in my tests.
During development we decided it was best to use Unity's physics solver so that our tools are fully compatible with Rigidbodies. Unfortunately, this means that drag can become unstable when the drag force and the weight of an object are close in magnitude.
A rough estimate for instability is to compute the value of:
4m / (p S Cd v)
(m is mass of an object, rho (p) is the density of the fluid (1.23), S is the largest area on the object, Cd is drag coefficient (in this case use the max of 1.2), v is an expected velocity for the object). If that value is less than the fixed delta time in your simulation, then you will likely have instability and objects will shoot off. The cube experiment for example gives a value of 0.545 so a simulation of that cube falling at 6.33 m/s should be stable with a time step of around 0.5 seconds - though I would advise using a much smaller one!
You can ensure stability by either: reducing the fixed delta time, increasing the mass of your objects, or reducing their areas. We have found a good value for fixed delta time to be 0.002 seconds.
Hope that helps!