That's a good alternative solution. But we would like to avoid the 30 seconds delay. We don't mind using 1 port bit to toggle (two bits are unnecessary). Also, we don't want to waste the high load driver (port B) for it. If we have time, we would try to recompile the bootloader. But we can't afford too much time dealing with it now, since this is just a side project.
But the way, we use bootloader because we want to disable the Jtag during runtime, so we can have 4 more A2D channels (port A sharing with Jtag). Unfortunately, we are concerned that once we disable jtag, it would be tough to get it back.