gil not hold during ~scoped_interpreter()

[iMerope]

program sometimes core dump during destruction of scoped_interpreter，the error is as following:

terminate called after throwing an instance of 'std::runtime_error'
what(): pybind11::object_api<>::operator() PyGILState_Check() failure.
(the above message is from cast.h)

after checking the backtrace:
Program terminated with signal SIGABRT, Aborted.
#0 __GI_raise (sig=sig@entry=6) at ../sysdeps/unix/sysv/linux/raise.c:54
54 return INLINE_SYSCALL (tgkill, 3, pid, selftid, sig);
[Current thread is 1 (Thread 0x7f3ff7fff700 (LWP 8394))]
(gdb) bt
#0 __GI_raise (sig=sig@entry=6) at ../sysdeps/unix/sysv/linux/raise.c:54
#1 0x00007f4023a00a8a in __GI_abort () at abort.c:89
#2 0x00007f4024336637 in __gnu_cxx::__verbose_terminate_handler () at ../../.././libstdc++-v3/libsupc++/vterminate.cc:95
#3 0x00007f40243420d6 in __cxxabiv1::__terminate (handler=)
at ../../.././libstdc++-v3/libsupc++/eh_terminate.cc:47
#4 0x00007f40243411d9 in __cxa_call_terminate (ue_header=ue_header@entry=0x7f3ff0009d30)
at ../../.././libstdc++-v3/libsupc++/eh_call.cc:54
#5 0x00007f4024341b04 in __cxxabiv1::__gxx_personality_v0 (version=, actions=6,
exception_class=5138137972254386944, ue_header=0x7f3ff0009d30, context=)
at ../../.././libstdc++-v3/libsupc++/eh_personality.cc:677
#6 0x00007f4023d928a3 in ?? () from /lib64/libgcc_s.so.1
#7 0x00007f4023d92dd7 in _Unwind_Resume () from /lib64/libgcc_s.so.1
#8 0x00007f40270df6c6 in pybind11::finalize_interpreter ()

I think the error is from here:

inline void finalize_interpreter ();

handle builtins(PyEval_GetBuiltins());
const char *id = PYBIND11_INTERNALS_ID;

// Get the internals pointer (without creating it if it doesn't exist).  It's possible for the
// internals to be created during Py_Finalize() (e.g. if a py::capsule calls `get_internals()`
// during destruction), so we get the pointer-pointer here and check it after Py_Finalize().
detail::internals **internals_ptr_ptr = detail::get_internals_pp();
// It could also be stashed in builtins, so look there too:

**/******************* calls object_api<Derived>::operator() but  PyGILState_Check() failure********************/**

if (builtins.contains(id) && isinstance<capsule>(builtins[id]))
    internals_ptr_ptr = capsule(builtins[id]);

Py_Finalize();

if (internals_ptr_ptr) {
    delete *internals_ptr_ptr;
    *internals_ptr_ptr = nullptr;
}

the simple structure of my program is like this:
scoped_interptreter is a local static variable in shared.so，gil should be always hold by main thread, but somehow lost during finalizing?

// share.so
int InitScopedInterpreter()
{
static py::scoped_interpreter guard();
}

// main
int main(){
InitScopedInterpreter();
PyEval_InitThreads();
}

I tried to add PyGILState_Ensure() at the end of main(), but not work, Then I tried to add it in finalize_interpreter(), it works.

my question is:

1. why finalize_interpreter failed? because gil was accidently lost?
2. how to ensure gil is been hold during finalize_interpreter?