> In your code, huge data passed to .write(huge) may be 
> remained in the internal buffer.

If you mean the buffered writer, then I don't see the problem. A large bytes object in pending_bytes gets temporarily referenced by _textiowrapper_writeflush(), and self->pending_bytes is cleared. If the buffer's write() method fails, then the bytes object is simply deallocated.

If you mean pending_bytes in the text wrapper, then I also don't see the problem. It always gets flushed if pending_bytes_count exceeds the chunk size. If pending_bytes is a list, it never exceeds the chunk size. It gets pre-flushed to avoid growing beyond the chunk size. If pending_bytes isn't a list, then it can only exceed the chunk size if it's a bytes object -- never for an ASCII str() object. _textiowrapper_writeflush() does not call PyBytes_AsStringAndSize() for that case.