Meta-weighing is weighing the weighing: when both teams have made impact comparisons, the round needs an argument about which comparison method itself should win. If you weigh on probability and they weigh on magnitude, meta-weighing argues "prefer probability, because rewarding speculative magnitude claims incentivizes fearmongering and makes debates unresolvable," or the reverse.
It becomes necessary precisely at higher levels, where both teams weigh competently and the mechanisms genuinely conflict; without meta-weighing, the judge chooses between frameworks on their own instincts, which is the definition of losing control of the round. Execution requires an actual warrant for the priority claim, not repetition of your mechanism at higher volume. Timing matters too: introduce it in summary so it can be developed, rather than springing it in final focus where it arrives too late to be credible. Close elimination rounds between strong teams are very often decided exactly here.