Traditionally if you update the contents of a running binary in-place in Unix you're liable to get a SIGBUS. The way to do it is unlink the existing binary and create a new one with the same name. Which is eg what install(1) does.
That isn't robust to failures, though. What you sould be doing is first writing the new binary using a temporary file name, and then you can rename the new file to the correct binary name, which is an atomic operation.
