Dynamic throttling thresholds for Exchange Online EWS/REST connections per mailbox

Answer

The per‑mailbox soft limit of ~27‑30 concurrent EWS/REST connections is not fixed; Microsoft dynamically adjusts this threshold based on tenant‑wide conditions. When the effective limit drops below the published value, clients receive HTTP 429 with a Retry‑After header and an error category of Throttling (MailboxThrottled or ConnectionThrottled).

Confirmed facts

• Microsoft documents a soft limit of ~27‑30 concurrent connections per mailbox and returns HTTP 429 with Retry‑After when exceeded.
• Throttling policies are system‑managed; Get-ThrottlingPolicy shows the default policy but does not expose per‑mailbox editable limits without creating a custom policy.
• Exceeding the dynamic limit yields ErrorCategory=Throttling and the response includes Retry‑After.

Likely explanation (based on observable behavior)

The adjustment appears to be driven by aggregate signals that reflect overall service pressure:

• Tenant‑level request volume – high overall EWS/REST traffic across all mailboxes can cause the per‑mailbox ceiling to be lowered.
• Mailbox‑specific activity – recent burst patterns, large mailbox size, or intensive sync operations raise the mailbox’s weight in the calculation.
• Service health indicators** – ongoing incidents, degraded components, or maintenance windows trigger more conservative limits.
• Licensing tier** – mailboxes assigned to higher‑tier plans (e.g., E5) may receive a larger share of the dynamic budget compared to lower tiers.

These factors are not exposed via admin APIs, which is why the exact formula remains undisclosed.

How to distinguish dynamic‑limit throttling from the published soft limit

1. Capture the full 429 response, noting the Retry‑After value and the exact ErrorCategory (MailboxThrottled vs ConnectionThrottled).
2. Compare the observed concurrent connection count (from client‑side pooling logs) with the baseline 27‑30 range.
   • If the count is ≤ 27‑30 and you still see 429, the limit has been dynamically lowered.
   • If the count exceeds 27‑30, the throttling aligns with the published soft limit.
3. Check tenant‑wide health via the Microsoft 365 admin center Service health page; any ongoing incidents correlate with dynamic limit reductions.
4. Review mailbox usage with Get-MailboxStatistics (look for large ItemCount or recent high‑volume operations) to see if the mailbox itself is a contributing factor.
5. Optionally, create a custom throttling policy and assign it to a test mailbox; if the 429 behavior changes, you confirm that the default system policy is being dynamically adjusted.

Missing diagnostic detail

To refine the recommendation, please confirm the Retry‑After you are observing in the 429 responses (e.g., 30 seconds, 2 minutes). This value indicates whether the throttling is a short‑lived burst‑control action or a longer‑cooldown caused by a tenant‑wide limit adjustment.