Dynamic throttling thresholds for Exchange Online EWS/REST connections per mailbox
28K reputation · 21 Nov 2023, 21:58 UTC
Goal: Determine what factors cause the Exchange Online EWS/REST connection limit per mailbox to shift from the documented soft limit of ~27‑30 concurrent connections under varying tenant load.
Context: Microsoft publishes a soft limit and returns HTTP 429 with Retry‑After when exceeded, but the internal algorithm that adjusts this limit dynamically based on overall tenant usage is not disclosed. This creates uncertainty for applications that rely on connection pooling and may experience intermittent pool exhaustion even when observed concurrency stays below the published threshold.
Uncertainty: Without visibility into the tenant‑wide signals (e.g., overall mailbox activity, service health, or licensing tier) that trigger the adjustment, it is unclear how to predict or monitor when the effective limit will be lower than the published value.
Questions:
- What tenant‑level metrics or service conditions influence the dynamic adjustment of the EWS/REST connection limit per mailbox?
- How can administrators distinguish throttling caused by the undisclosed dynamic limit from throttling that aligns with the published soft limit?
