What is the correct way to handle 429s on Predictive Routing API?

Context:
Running JMeter 5.6.2 from Singapore. Targeting POST /api/v2/predictiverouting/outboundcampaigns. At 15 concurrent threads, I get consistent 429 Too Many Requests errors. Single thread works fine. My current script just retries immediately on failure, which seems to make it worse.

Question:
What is the correct way to implement retry logic or pacing in JMeter to stay within Genesys Cloud rate limits for this endpoint? Is there a specific header I should check?