With this change I do see mass conservation on domains with curved boundaries with edges.
We have mass conservation up to floating point precision with SSP RK3.
With ERK3 is do observe a relative loss of mass of 10^-9 for a coarse mesh with curved boundaries. I am not exactly sure where this is coming from - but it might be due to accumulation of round-off errors in the more aggressive ERK3 flux reconstruction.
With this change I do see mass conservation on domains with curved boundaries with edges.
We have mass conservation up to floating point precision with SSP RK3.
With ERK3 is do observe a relative loss of mass of 10^-9 for a coarse mesh with curved boundaries. I am not exactly sure where this is coming from - but it might be due to accumulation of round-off errors in the more aggressive ERK3 flux reconstruction.