Run failure in MVK test with F2010 compset, ne4_oQU240 grid

[mkstratos]

The MVK test is failing due to a runtime issue:

759:  ERROR: shr_reprosum_calc ERROR: NaNs or INFs in input
762: Image              PC                Routine            Line        Source             
762: libpnetcdf.so.3.0  0000155554F1412A  tracebackqq_          Unknown  Unknown
762: e3sm.exe           0000000004E355D0  shr_abort_mod_mp_         114  shr_abort_mod.F90
762: e3sm.exe           0000000004F25355  shr_reprosum_mod_         644  shr_reprosum_mod.F90
762: e3sm.exe           00000000025BF9BF  compose_repro_sum         436  compose_mod.F90
762: e3sm.exe           0000000001D97A69  Unknown               Unknown  Unknown
762: e3sm.exe           0000000001DAB18E  Unknown               Unknown  Unknown
762: e3sm.exe           0000000001DDC494  Unknown               Unknown  Unknown
762: e3sm.exe           0000000001D86FF0  Unknown               Unknown  Unknown
762: e3sm.exe           0000000001D0651A  sl_advection_mp_p         258  sl_advection.F90
762: e3sm.exe           0000000001CB663A  prim_driver_base_        1428  prim_driver_base.F90
762: e3sm.exe           00000000025A36F2  dyn_comp_mp_dyn_r         401  dyn_comp.F90
762: e3sm.exe           0000000001B7BA43  stepon_mp_stepon_         582  stepon.F90
762: e3sm.exe           00000000005E8F44  cam_comp_mp_cam_r         352  cam_comp.F90
762: e3sm.exe           00000000005CF055  atm_comp_mct_mp_a         583  atm_comp_mct.F90
762: e3sm.exe           00000000004558BB  component_mod_mp_         757  component_mod.F90
762: e3sm.exe           000000000043250D  cime_comp_mod_mp_        3116  cime_comp_mod.F90
762: e3sm.exe           00000000004554F0  MAIN__                    153  cime_driver.F90
762: e3sm.exe           0000000000426422  Unknown               Unknown  Unknown
762: libc-2.28.so       000015554503D493  __libc_start_main     Unknown  Unknown
762: e3sm.exe           000000000042632E  Unknown               Unknown  Unknown

This is with the F2010 compset on ne4_oQU240 grid. Judging on the last few days of fails, it's not deterministic at what time step and which ensemble member presents the failure, but it's failing every time.
It may be related to PR #6054, as it's able to run at commit 876a76, but not any subsequent commits.