the middle layer velocity blows up

[cheng_dada]

I update the ROMS from Version 3.4 to 4.1, then the middle layer velocity blows up in the 9000th step, while it runs well in ROMS version 3.4 with same input nc and initial conditions. Fig.1 is the blow up point, it is on the shore of an island.
I try to make the depth get more smooth, or change it into a dry point, but it will still blow up. Hope to get some help with this problem. Great thanks!

Fig.1

--------------------------------------------------------------------------------------------------------------------------------------------------------
Here is my cpp define:

#define SOLVE3D
#define UV_ADV
#define UV_COR
#define UV_QDRAG
#define DJ_GRADPS

/************ Grid configuration */
#define MASKING
#define SPHERICAL
#define CURVGRID

/**** Mixing ********************/
#define TS_DIF2
#define MIX_GEO_TS
#define UV_VIS2
#define MIX_S_UV

#define LMD_MIXING

# ifdef MY25_MIXING
# define KANTHA_CLAYSON
# define N2S2_HORAVG
# define RI_SPLINES
# endif

# ifdef LMD_MIXING
# define LMD_RIMIX
# define LMD_CONVEC
# define LMD_SKPP
# define LMD_BKPP
# define LMD_NONLOCAL
# define RI_SPLINES
# endif

/**** Equation of State *********/
#define NONLIN_EOS
#define SPLINES_VDIFF
#define SPLINES_VVISC
#define SALINITY

/*************** Surface Forcing */
# define SOLAR_SOURCE
# define QCORRECTION
# define DIURNAL_SRFLUX

/************** Other options ***/
#define RADIATION_2D
#define AVERAGES
#define ANA_BTFLUX
#define ANA_BSFLUX
#define LIMIT_BSTRESS
#define LIMIT_VVISC
#define LIMIT_VDIFF

/******** Tide ************/
# define RAMP_TIDES
# define SSH_TIDES
# define UV_TIDES
# define ADD_M2OBC
# define ADD_FSOBC

/* nudging the time in *.in shall be set */
#undef OUT_DOUBLE
#define RST_SINGLE

[wilkin]

You chose 4-th order centered advection for tracers. Is that what you used previously with the stable model?

Your barotropic Courant number is 0.6 which is getting a bit large, though still stable. You might need to increase NDTFAST.