Does GitHub's internal request queuing cause latency spikes only under concurrent authenticated API calls?

The goal is to determine whether the latency increases observed when sending multiple concurrent requests to GitHub's REST API stem from internal request queuing mechanisms rather than from hitting the documented primary or secondary rate limits.

Because GitHub's public documentation notes that response times may vary under high concurrency but does not specify the exact throttling behavior, it remains unclear at what concurrency level internal queuing begins to add measurable delay, and how that interacts with the X‑RateLimit‑Remaining and X‑RateLimit‑Reset headers. What concurrency level triggers a noticeable latency increase? How do the rate‑limit headers correlate with observed latency spikes? Are latency spikes present for both REST endpoints and GraphQL batch queries under the same concurrency conditions?