Viscosity in 2D_cavity

In 2D_cavity, we select constant alpha viscosity with a 0 value, which should raise a 0 field viscosity in a perfect world.

On some platform (Irène/gnu11) we get:

66: Field 'viscosity':	[avg:8.65548e-312] max abs diff: 1.12532e-311 max relative diff: 2  (> 1e-10) 

Which is basically noise around zero, which leads to important relative differences.

Are we happy with that ? isn't there a simpler way to specify a viscosity of 0 ?

Or should I just relax the relative difference and/or add the possibility to specify a noise limit to says "this is 0, really" ?