LinkedIn Marketing API: Concurrency limit per access token unclear

Goal

Determine the maximum number of concurrent requests that a single LinkedIn access token can sustain without significant latency increase or rate‑limit errors.

Constraints

LinkedIn documentation lists per‑app and per‑call limits but omits a concurrent‑connection ceiling. Observed latency spikes and 429 responses appear only when multiple threads share the same token, with no Retry‑After header.

Unresolved Questions

• What is the documented or implicit concurrency limit for a single access token on the Marketing API?
• Does LinkedIn enforce a per‑token queue that triggers latency spikes before returning 429?
• How can a client detect and adapt to this undocumented throttling in real time?