If your SQL Server estate runs linked server queries and installed the October 13 update, connections to external databases may pile up and performance may sag over time. Microsoft has confirmed the problem. It offers manual cleanup and a service restart as stopgaps, and a mitigation exists through Microsoft support for business customers. An out-of-band fix is due within days. Database and operations teams should measure their exposure now rather than wait for the slowdown to find them.

What happened

According to The Register, Microsoft has acknowledged that the October 13 update (KB5126646) changed behavior so that connections opened by linked server queries may stay open after the query finishes. Microsoft says connections to external databases can accumulate as more queries run, and that this can slow or degrade server performance after prolonged run time.

The problem touches linked servers that use OLE DB providers. Microsoft named PostgreSQL and Oracle as connections known to be affected. The Register reports that the issue spans Windows versions back to Windows 10 1809 and LTSC 2019, and Windows Server from 2025 back to 2012, with Windows 11 26H2 the exception. BleepingComputer’s coverage says Microsoft links SQL Server performance and stability problems to Windows updates released since the October 2026 Patch Tuesday.

Microsoft’s interim advice is to close idle connections by hand where admins can identify them, or to restart the SQL service. It warns that both steps can interrupt active queries or transactions and that the problem can return after enough run time. The Register adds that a mitigation is available through Microsoft support for business, and that an out-of-band update should arrive within the next few days.

Why it matters

Connection leaks rarely fail loudly. They build slowly, so a server can look healthy for days and then degrade under load. That makes the cause easy to misdiagnose as a workload or network issue. Linked servers also serve as a common glue layer between SQL Server and other databases, so the slowdown can reach reporting jobs, integrations and anything else that crosses that boundary.

Separately, The Register notes that the October 2026 Windows Server update left some Exchange servers unstable until admins applied a registry edit or the November update fixed it. That is a different update and a different product, and a fix already exists. Still, two Microsoft update problems in the same period suggest that installing patches on release day carries operational risk, so a staged rollout deserves a place in your process.

What it means for you

Start by finding out whether you are exposed. Inventory which hosts installed KB5126646 and which of them run linked servers through OLE DB providers, especially to PostgreSQL or Oracle. Then check open connection counts on the external databases and compare them with your pre-patch baseline. A steady upward drift is the signal.

Next, set an alert on connection counts at the remote end, where limits can be hit first. If you must restart the SQL service, schedule it in a maintenance window. Where you run availability groups or clusters, treat the restart as a controlled failover rather than a surprise. Microsoft warns that restarts can cut active transactions.

If you hold a business support contract with Microsoft, ask about the mitigation now instead of waiting for the out-of-band release. Microsoft expects that release within days, but a date is not guaranteed, so plan for a longer wait. For hosts that have not yet installed the update, consider holding it on systems that depend heavily on linked servers until the fix is out and tested in staging. Teams that track which tools are growing or fading in their stack can use our Stack Graveyard to judge how far to lean on linked-server patterns long term.

What to watch

  • Release of the out-of-band update Microsoft says should arrive within days, and whether it fully stops the leak.
  • Any expansion of the list of affected providers beyond PostgreSQL and Oracle.
  • Whether the fix rolls into the regular cumulative update or needs separate deployment.
  • Whether Microsoft confirms further regressions from updates in the same patch cycle.

More analysis for operators sits on our Enterprise and SaaS page.

Sources