Use cbt results when deciding to launch parallel introduction circuit?

Right now, circuituse.c has this lovely snippet in circuit_get_best:

/* XXX022 make this 15 be a function of circuit finishing times we've
 * seen lately, a la Fallon Chen's GSoC work -RD */
#define REND_PARALLEL_INTRO_DELAY 15
    if (purpose == CIRCUIT_PURPOSE_C_INTRODUCE_ACK_WAIT &&
             !must_be_open && circ->state != CIRCUIT_STATE_OPEN &&
             circ->timestamp_created + REND_PARALLEL_INTRO_DELAY < now) {
      intro_going_on_but_too_old = 1;
      continue;
    }

Now that we've got our circ_times stuff, we can totally get a better estimate of "that time after which we should launch another introduction circuit but still give this one time to finish." Yes?

Parenthetically, I have no idea what "!must_be_open" is doing there.