I am running an idealized simulation of a flow in a periodic channel with 2.5km resolution. At the surface, it is forced by relaxation to a zonally uniform temperature profile. After less than 2 years, the simulation develops huge SST values (100 deg) and blows up. I verified that the surface flux is what it is supposed to be (given these crazy SST values): enormous amounts of heat are lost in places where SST is huge.
Here are the options I am using. All ANA fluxes (surface and bottom) are set to 0. I also tried using several advection schemes, setting nudging coefficients to 0 (eventually, I want to use sponge layers in the north/south) and turning LMD_MIXING on and off:
#define UV_ADV
#define UV_COR
#define UV_QDRAG
#define DJ_GRADPS
#define UV_VIS2
#define MIX_GEO_UV
#define MIX_GEO_TS
#define SPLINES_VDIFF
#define SPLINES_VVISC
#define TS_DIF2
#define SOLVE3D
#define SALINITY
#define NONLIN_EOS
#define SRELAXATION
#define QCORRECTION
#define LMD_MIXING
#ifdef LMD_MIXING
# define LMD_RIMIX
# define LMD_CONVEC
# define LMD_SKPP
# define LMD_NONLOCAL
# define RI_SPLINES
#endif
#define ANA_GRID
#define ANA_SST
#define ANA_DQDSST
#define ANA_SSS
#define ANA_SMFLUX
#define ANA_STFLUX
#define ANA_SRFLUX
#define ANA_BTFLUX
#define ANA_BSFLUX
#define ANA_TCLIMA
#define ANA_INITIAL
#define ANA_NUDGCOEF
#define STRONG_EDDIES