Skip to content

Est. 2011ยทMicrosoft Partner ID 7033487ยทDelivery under 3 minยทSupport 7 days a week

Your vault is empty.

License Error Event ID 9646

Event ID 9646: a client blew past the Exchange session or object limit

11 min read Updated October 5, 2026 Exchange Server

Fix it now

The information store cut off a session that asked for more objects than its budget allows, and the event names the object type and the account. Users see it as “Your server administrator has limited the number of items you can open simultaneously”. It is a badly behaved client or an overloaded public folder mailbox far more often than it is a sign the server is undersized.

For the public folder case, run these in Exchange Online PowerShell or the Exchange Management Shell

Get-Mailbox | Where-Object DefaultPublicFolderMailbox -ne $null | Format-Table Name,*PublicFolderMailbox*
New-Mailbox -Name "<new public folder mailbox>" -PublicFolder
Update-PublicFolderMailbox -InvokeSynchronizer -Identity "<new public folder mailbox>"
Set-Mailbox -Identity "<user>" -DefaultPublicFolderMailbox "<new public folder mailbox>"
  1. Open the event in Windows Logs then Application and read two things: the object type named, and the account or session it belongs to.
  2. If users hit this expanding the public folder tree, update Outlook first. Microsoft published fixes for Outlook 2016 and Outlook 2013 in the April 2016 updates specifically for this symptom.
  3. If the account is a service account, identify the product behind it: archiving and journaling tools, CRM connectors, backup agents and home-grown integrations are the usual four.
  4. Stop that service or disable that add-in for an hour and see whether the events stop. Proving causation this way is faster and far safer than editing store limits on a hunch.
  5. Where the default public folder mailbox is exhausted rather than any one client misbehaving, add another public folder mailbox and move users onto it, or clear their pin with Set-Mailbox <user> -DefaultPublicFolderMailbox $null so automatic distribution includes the new one.

Microsoft does not currently publish the store’s per-session object limits, their registry locations or their defaults. If a vendor tells you their product needs a specific value, get the value and the path from them in writing rather than from a search result.

If the events stop after the client change, you are done. If they continue, the next section covers what the store is counting and why it stops.

Why it happens

A MAPI session is a conversation between a client and a mailbox. During that conversation the client opens objects: folders it is browsing, messages it is reading, attachments it is saving, views it has built. Well-behaved clients open what they need and release it. The store keeps a per-session count for each object type and enforces a ceiling on each, so one runaway conversation cannot exhaust a shared server. A separate ceiling governs how many sessions a single user may hold at once.

The object type named in the event is the most useful thing in it. A folder count blown out points at something enumerating the whole mailbox, typically an indexing, archiving or synchronisation tool. A message count points at something opening items in bulk without closing them. A session count points at an application opening a new session for every operation instead of reusing one, which is a classic mistake in home-grown integrations, or at a person with a great many delegated mailboxes open at once.

The same limit reaches applications as 0x80040112, which the Exchange protocol documentation publishes as TooManySessions: a server or service is unable to create any more sessions. That is worth knowing because the code is frequently misread as a memory problem on the workstation. It is not. It is a refusal from the server, and it means the same thing the event does.

The public folder hierarchy has exhausted the mailbox serving it

You have this one if Users see the message when expanding the public folder tree in Outlook or Outlook on the web, and it affects several people rather than one.

  1. Install the April 2016 updates for Outlook 2016 or Outlook 2013 if the affected clients are on older builds. Microsoft published them for this symptom specifically.
  2. Create another public folder mailbox with New-Mailbox -Name "<name>" -PublicFolder and give it the hierarchy with Update-PublicFolderMailbox -InvokeSynchronizer -Identity "<name>".
  3. Find the users pinned to the overloaded one: Get-Mailbox | Where-Object DefaultPublicFolderMailbox -like "*<name>*".
  4. Move them, or clear the pin with Set-Mailbox <user> -DefaultPublicFolderMailbox $null so they are distributed across every available mailbox including the new one.

This is the documented cause where the threshold for automatically provisioning a new public folder mailbox has not been reached but the existing one is already saturated.

An add-in or third-party application is leaking objects

You have this one if The same account appears in every event, and it belongs to a service or to a user with one specific add-in installed.

  1. Identify the product and check whether the vendor has a newer build. Object leaks are common and commonly fixed.
  2. Disable the add-in or stop the service temporarily and confirm the events stop.
  3. Where the application polls, reduce its frequency and its concurrency before considering anything else.
  4. Raise a limit only if the vendor confirms the behaviour is expected rather than a defect, and only to the figure they document.

A user has too many mailboxes open at once

You have this one if A session count rather than an object count, and the account belongs to a person rather than a service.

  1. Ask what the user has open. Delegated mailboxes, shared mailboxes and additional accounts each consume sessions.
  2. Move genuinely shared workloads into shared mailboxes accessed with full access rather than as extra accounts in the profile.
  3. Where a whole team opens the same mailbox, review whether every member needs it permanently open.

The server is genuinely overloaded rather than abused

You have this one if RPC server too busy results accompany the events, and the pressure comes from many clients rather than one.

  1. Check disk latency on the database volumes before anything else. Slow storage makes sessions live longer and pile up.
  2. Review how many active mailboxes the server carries against what the hardware was sized for.
  3. Spread mailboxes across more databases or more servers rather than raising per-session limits.
  4. Treat limit changes as the last resort, not the first response.

Full reference

Reading the event to find the culprit

Object type named What usually causes it
Folders A tool enumerating the whole mailbox: indexing, archiving, migration or synchronisation
Messages An application opening items in bulk and not releasing them
Sessions, for a person Many shared or delegated mailboxes open, or a client reconnecting repeatedly
Sessions, for a service account An integration opening a new session per operation instead of reusing one
Any type, for many users at once, in public folders The public folder mailbox serving the hierarchy is saturated

What is and is not published

Microsoft publishes the client-facing message and the public folder remedy, and the Exchange protocol documentation publishes the session-limit error code. It does not currently publish the store’s per-session object ceilings, the registry location that holds them, or their default values. That matters, because a great deal of the advice in circulation for this event consists of registry paths copied between forum posts, applied on a server, and followed by an Information Store restart that dismounts every database.

If you conclude a limit genuinely has to move, get the value name and the number from the vendor whose product needs it, in their own documentation, and record the change and its reason where the next administrator will find it. A non-default limit nobody can explain is worse than the event it silenced.

Diagnosing without changing anything

  1. Note the account and object type from several events, not one.
  2. Establish whether the accounts share a client, an add-in or a public folder mailbox.
  3. Stop the suspected component for a defined window and watch the event log.
  4. If the events stop, you have your answer. If they do not, you have eliminated it cheaply.
  5. Only after that consider whether the server is carrying more load than it was sized for.

Why restarting the store is not a fix

Restarting the Information Store clears the sessions and buys quiet, at the cost of dismounting every database and disconnecting every user on that server. The client that leaked objects will do it again, usually within the hour. Use the quiet period to identify the client rather than treating the restart as the remedy, and do not restart on a hunch that the limits are wrong.

Related symptoms worth checking at the same time

  • RPC server too busy results, which frequently travel with this event and point at load rather than at one client.
  • Users on the same public folder mailbox reporting slow tree expansion rather than an outright error.
  • Disk latency on the database volumes, which lengthens every session and makes a marginal limit into a breached one.
  • Mobile clients reconnecting in a loop, which consumes sessions faster than any add-in.
  • Whether the accounts in the events are all on one database or one server, which separates a client problem from a capacity one.

When a licence is the actual fix

Fixing this usually costs nothing, and in most cases raising a limit is the wrong answer anyway because the real fault is a client or an add-in. Adding a public folder mailbox costs nothing either. Licensing enters the conversation only if the honest conclusion is that this server carries more mailboxes than it should and needs its load spread across more databases or another server. Standard Edition supports five mounted databases and Exchange Server SE Enterprise supports up to 100, and a second server needs its own licence and its own client access licences. Fix the client first. If the capacity problem survives that, we can price either route once you know which one you actually have.

Every code this article covers

Code What it points at Source
Event ID 9646 The store refused a session further requests because it exceeded the ceiling for one class of object. No current vendor description is published; the client-facing message is ‘Your server administrator has limited the number of items you can open simultaneously’ not published by the vendor
MapiExceptionSessionLimit The same refusal surfaced to an application as a MAPI exception not published by the vendor
MapiExceptionRpcServerTooBusy The store shed work because it was under too much load to accept more not published by the vendor
0x80040112 TooManySessions (MAPI_E_SESSION_LIMIT): a server or service is unable to create any more sessions. It is a server-side refusal, not a workstation memory problem Microsoft Learn

Confirm the fix worked

  1. The application log runs a full working day without event 9646 for the same account.
  2. The client or service you changed is functioning normally rather than silently failing.
  3. Users can expand the public folder tree from a client that previously could not.
  4. RPC server too busy results have also stopped, since they often travel with this event.
  5. Any limit change you made is recorded with the vendor documentation that justified it.

Questions people ask about this

Is it safe to raise the limits?

Changing a documented value for a documented reason is supportable. The problem is that Microsoft does not currently publish these ceilings or their registry locations, so most of the advice in circulation is unverified, and applying it costs an Information Store restart that dismounts every database.

How do I know which user is responsible?

The event identifies the session and account. Correlate that with what the account runs: a person with an add-in, or a service account belonging to a product. If several accounts appear, look for a common client rather than a common user.

What does 0x80040112 actually mean?

The Exchange protocol documentation publishes it as TooManySessions: a server or service is unable to create any more sessions. It is the server refusing, not the workstation running out of memory, which is how it is usually misread.

Do I need a bigger licence to fix this?

Almost never. This is nearly always a client or add-in defect, and both diagnosing and fixing it cost nothing. Licensing matters only if you conclude the server needs more databases or a second server to spread the load.

Will restarting the store fix it permanently?

No. It clears the sessions and buys quiet, but the client that leaked objects will do it again. Use the pause to identify the client.

Related error codes

Was this article helpful?

Your feedback helps us improve our documentation.

Related articles

Free Fix 0x8004010F: Outlook cannot download the offline address book from Exchange Free Fix 451 4.4.397 and 451 4.4.316: queues stuck in Retry to external hosts Free Fix Error 1618 and a half finished Exchange CU: setup will not start again Free Fix 0x80072030 during Exchange setup: Active Directory objects cannot be found
โ† Back to Knowledge Base