Unlock fails with STATUS_LOGON_FAILURE / STATUS_INTERNAL_ERROR (0xC000006D / 0xC00000E5) when the workstation is locked while Group Policy registry processing temporarily removes HideFastUserSwitching

CyberKDH 0 Reputation points
2026-10-28T10:03:58.4133333+00:00
  1. Environment

ItemValueOSWindows 11 Version 24H2 (Korean), x64, OS build 26100.1742OSWindows 11 Version 24H2 (Korean), x64, OS build 26100.1742PlatformVMware Workstation virtual machine (also observed on a physical enterprise PC)DomainActive Directory domain joinedPolicyComputer Configuration > Policies > Administrative Templates > System > Logon > Hide entry points for Fast User Switching = Enabled (HKLM\SOFTWARE\Microsoft\Windows\CurrentVersion\Policies\System\HideFastUserSwitching = 1)Credential provider used for the failing sign-inBuilt-in Password Credential Provider {60B78E88-EAD8-445C-9CFD-0B87F74EA6CD}Third-party credential providers / filtersNone registered at the time of the clean reproduction (section 6.2). The only registered filter was {DDC0EED2-ADBE-40b6-A217-EDE16A79A0DE}. The same failure was also seen earlier with a third-party credential provider installed## 3. Summary

When Group Policy is applied, the Registry client-side extension deletes and rewrites Policies\System\HideFastUserSwitching. We measured the value missing for 9 to 33 ms during gpupdate /target:computer /force.

If the workstation is locked during that window, LogonUI starts the credential providers with CPUS_LOGON instead of CPUS_UNLOCK_WORKSTATION, even though the same user's session is locked on the console. The value is then restored to 1. When the user submits valid credentials on that screen, Winlogon rejects them:

  • with 0xC000006D (STATUS_LOGON_FAILURE) / substatus 0xC00000E5 (STATUS_INTERNAL_ERROR)
  • without any LSA authentication (no 4624 / 4625 / 4648 for the submission in the Security log)

The built-in Password credential provider shows "The password is incorrect. Try again." although the password is correct.

Retrying does not help: the screen stays in CPUS_LOGON. The state recovers only when the user dismisses the error dialog and then stops interacting. While the error dialog is open, nothing happens. About 30 seconds after the dialog is dismissed, LogonUI returns to the lock screen curtain and re-enumerates the providers with CPUS_UNLOCK_WORKSTATION. The next attempt then succeeds as a normal unlock (logon type 7).

In an enterprise environment with 31 GPOs and computer policy refreshed every 30 to 60 minutes, users hit this repeatedly (8 consecutive failures in one case).

4. Steps to reproduce

4.1 Deterministic reproduction (simulates the Registry CSE gap)

  1. Domain-joined Windows 11 24H2 with "Hide entry points for Fast User Switching" = Enabled (value 1).
  2. Sign in as a domain user.
  3. Run the script below as Administrator in the user's desktop session.
    • It deletes the value, locks immediately, and restores the value to 1 after 3 seconds, which is what the Registry CSE does, only with a longer gap.
    1. Wait more than 3 seconds on the lock screen, dismiss the curtain, choose the Password sign-in option and enter the correct password.
# repro.ps1  - run as administrator in the interactive user session

4.2 Measuring the natural gap

While polling the value every ~0.1 ms as Administrator, run gpupdate /target:computer /force three times:

RunValue removedValue rewrittenGap117:11:46.47517:11:46.50833 ms117:11:46.47517:11:46.50833 ms217:11:54.13417:11:54.14511 ms317:12:01.59217:12:01.6019 ms### 4.3 Natural reproduction (probabilistic)

Lock the workstation repeatedly while computer Group Policy is being processed (gpupdate /target:computer /force in a loop, or during background refresh). The failure occurs whenever the lock falls inside the gap.

5. Expected result

The locked session is unlocked (logon type 7), as it is when the value is present at lock time.

6. Actual result

ObservationDetailScenario given to credential providersCPUS_LOGON (1) at lock time, the console session belongs to the same locked user, HideFastUserSwitching absent at that momentScenario given to credential providersCPUS_LOGON (1) at lock time, the console session belongs to the same locked user, HideFastUserSwitching absent at that momentSubmission result0xC000006D / 0xC00000E5Message (built-in Password CP)"암호가 올바르지 않습니다. 다시 시도하세요." ("The password is incorrect. Try again.")Security log for the submissionNo 4624 / 4625 / 4648 from Winlogon (svchost / lsass); the request never reaches LSARetry5 consecutive attempts with the built-in Password CP all failedEsc / Ctrl+Alt+Del after the failureNo changeRecoveryAbout 30 s after the error dialog is dismissed and no further input occurs, LogonUI returns to the curtain and re-enumerates providers with CPUS_UNLOCK_WORKSTATION (HideFastUserSwitching = 1); the next attempt succeeds (4624 logon type 7)Error dialog left openThe curtain does not return while the error dialog is shown (observed for more than 90 s); the idle timeout starts only after the dialog is dismissed### 6.1 Example timeline (one run, guest clock) TimeEvent17:55:58.124Value deleted, LockWorkStation17:55:58.124Value deleted, LockWorkStation17:55:58.128LogonUI started in the user's session (session 2)17:55:58.296Credential provider filter called with CPUS_LOGON, value absent17:56:01.176Value restored to 117:56:25Password CP sign-in fails ("The password is incorrect"), no LSA events17:56:52.953Same LogonUI process re-enumerates providers with CPUS_UNLOCK_WORKSTATION, value 117:56:55Curtain shown17:57:29.933 / 17:57:30.025Unlock succeeds: 4624 logon type 7 (svchost.exe, then lsass.exe)### 6.2 Clean reproduction (no third-party credential provider or filter registered) TimeEvent18:15:10.917Value deleted18:15:10.917Value deleted18:15:10.933LockWorkStation18:15:10.950LogonUI (pid 8664) running in session 2 (process start 18:15:10.936)18:15:13.942Value restored to 118:15:24Password CP sign-in with the correct password fails: "The password is incorrect. Try again."18:15:11 – 18:17:25No 4624 / 4625 / 4648 for the user in the Security log18:15:24 –Error dialog left open; the curtain does not return18:17:17After the error dialog is dismissed, the lock screen curtain is shown18:17:25.838 / 18:17:25.936Unlock succeeds: 4648 + 4624 logon type 7 (svchost.exe, then lsass.exe)### 6.3 Controls Value at lock timeValue at submissionResult11Success (normal unlock, CPUS_UNLOCK_WORKSTATION)11Success (normal unlock, CPUS_UNLOCK_WORKSTATION)00Success: CPUS_LOGON shown, Winlogon converts the logon into an unlock of the existing session (logon type 7)absent1Failure, 0xC000006D / 0xC00000E5, no LSA authentication## 7. Impact

  • Users in domains that enable "Hide entry points for Fast User Switching" can intermittently fail to unlock with correct credentials after a background Group Policy refresh.
  • The message suggests a wrong password, so users retry repeatedly or restart the PC.
  • Credential providers have no supported way to ask LogonUI to restart the usage scenario. ICredentialProviderEvents::CredentialsChanged re-enumerates credentials but does not call SetUsageScenario again, so the condition cannot be recovered from a provider.

8. Workaround

Dismiss the error dialog (OK), then do not interact with the sign-in screen for about 30 seconds. When the lock screen curtain returns, sign in again. Leaving the error dialog open does not recover.

9. Questions to Microsoft

  1. Is it expected that the Registry CSE removes and rewrites Policies\System values non-atomically during refresh, even when the resulting value does not change?
  2. Can Winlogon/LogonUI decide the unlock scenario independently of a transient missing HideFastUserSwitching value, or re-evaluate the scenario when the submission fails with STATUS_INTERNAL_ERROR?
  3. Is there a supported setting to avoid the gap (for example "Configure registry policy processing" options) without side effects on other registry-based policies?

    2. Environment

ItemValueOSWindows 11 Version 24H2 (Korean), x64, OS build 26100.1742PlatformVMware Workstation virtual machine (also observed on a physical enterprise PC)DomainActive Directory domain joinedPolicyComputer Configuration > Policies > Administrative Templates > System > Logon > Hide entry points for Fast User Switching = Enabled (HKLM\SOFTWARE\Microsoft\Windows\CurrentVersion\Policies\System\HideFastUserSwitching = 1)Credential provider used for the failing sign-inBuilt-in Password Credential Provider {60B78E88-EAD8-445C-9CFD-0B87F74EA6CD}Third-party credential providers / filtersNone registered at the time of the clean reproduction (section 6.2). The only registered filter was {DDC0EED2-ADBE-40b6-A217-EDE16A79A0DE}. The same failure was also seen earlier with a third-party credential provider installed ## 3. Summary

When Group Policy is applied, the Registry client-side extension deletes and rewrites Policies\System\HideFastUserSwitching. We measured the value missing for 9 to 33 ms during gpupdate /target:computer /force.

If the workstation is locked during that window, LogonUI starts the credential providers with CPUS_LOGON instead of CPUS_UNLOCK_WORKSTATION, even though the same user's session is locked on the console. The value is then restored to 1. When the user submits valid credentials on that screen, Winlogon rejects them:

  - with `0xC000006D` (STATUS_LOGON_FAILURE) / substatus `0xC00000E5` (STATUS_INTERNAL_ERROR)

  
     - **without any LSA authentication** (no 4624 / 4625 / 4648 for the submission in the Security log)
```   The built-in Password credential provider shows "The password is incorrect. Try again." although the password is correct.
   
   Retrying does not help: the screen stays in CPUS_LOGON. The state recovers only when the user dismisses the error dialog and then stops interacting. While the error dialog is open, nothing happens. About 30 seconds after the dialog is dismissed, LogonUI returns to the lock screen curtain and re-enumerates the providers with CPUS_UNLOCK_WORKSTATION. The next attempt then succeeds as a normal unlock (logon type 7).
   
   In an enterprise environment with 31 GPOs and computer policy refreshed every 30 to 60 minutes, users hit this repeatedly (8 consecutive failures in one case).
   
   ## 4. Steps to reproduce

   ### 4.1 Deterministic reproduction (simulates the Registry CSE gap)

```sql
  1. Domain-joined Windows 11 24H2 with "Hide entry points for Fast User Switching" = Enabled (value 1).

  
     1. Sign in as a domain user.

     
        1. Run the script below as Administrator in the user's desktop session.

        
              - It deletes the value, locks immediately, and restores the value to 1 after 3 seconds, which is what the Registry CSE does, only with a longer gap.

              
                 1. Wait more than 3 seconds on the lock screen, dismiss the curtain, choose the **Password** sign-in option and enter the correct password.
```   ```powershell
   # repro.ps1  - run as administrator in the interactive user session

4.2 Measuring the natural gap

While polling the value every ~0.1 ms as Administrator, run gpupdate /target:computer /force three times:

RunValue removedValue rewrittenGap117:11:46.47517:11:46.50833 ms217:11:54.13417:11:54.14511 ms317:12:01.59217:12:01.6019 ms ### 4.3 Natural reproduction (probabilistic)

Lock the workstation repeatedly while computer Group Policy is being processed (gpupdate /target:computer /force in a loop, or during background refresh). The failure occurs whenever the lock falls inside the gap.

5. Expected result

The locked session is unlocked (logon type 7), as it is when the value is present at lock time.

6. Actual result

ObservationDetailScenario given to credential providersCPUS_LOGON (1) at lock time, the console session belongs to the same locked user, HideFastUserSwitching absent at that momentSubmission result0xC000006D / 0xC00000E5Message (built-in Password CP)"암호가 올바르지 않습니다. 다시 시도하세요." ("The password is incorrect. Try again.")Security log for the submissionNo 4624 / 4625 / 4648 from Winlogon (svchost / lsass); the request never reaches LSARetry5 consecutive attempts with the built-in Password CP all failedEsc / Ctrl+Alt+Del after the failureNo changeRecoveryAbout 30 s after the error dialog is dismissed and no further input occurs, LogonUI returns to the curtain and re-enumerates providers with CPUS_UNLOCK_WORKSTATION (HideFastUserSwitching = 1); the next attempt succeeds (4624 logon type 7)Error dialog left openThe curtain does not return while the error dialog is shown (observed for more than 90 s); the idle timeout starts only after the dialog is dismissed ### 6.1 Example timeline (one run, guest clock) TimeEvent17:55:58.124Value deleted, LockWorkStation17:55:58.128LogonUI started in the user's session (session 2)17:55:58.296Credential provider filter called with CPUS_LOGON, value absent17:56:01.176Value restored to 117:56:25Password CP sign-in fails ("The password is incorrect"), no LSA events17:56:52.953Same LogonUI process re-enumerates providers with CPUS_UNLOCK_WORKSTATION, value 117:56:55Curtain shown17:57:29.933 / 17:57:30.025Unlock succeeds: 4624 logon type 7 (svchost.exe, then lsass.exe) ### 6.2 Clean reproduction (no third-party credential provider or filter registered) TimeEvent18:15:10.917Value deleted18:15:10.933LockWorkStation18:15:10.950LogonUI (pid 8664) running in session 2 (process start 18:15:10.936)18:15:13.942Value restored to 118:15:24Password CP sign-in with the correct password fails: "The password is incorrect. Try again."18:15:11 – 18:17:25No 4624 / 4625 / 4648 for the user in the Security log18:15:24 –Error dialog left open; the curtain does not return18:17:17After the error dialog is dismissed, the lock screen curtain is shown18:17:25.838 / 18:17:25.936Unlock succeeds: 4648 + 4624 logon type 7 (svchost.exe, then lsass.exe) ### 6.3 Controls Value at lock timeValue at submissionResult11Success (normal unlock, CPUS_UNLOCK_WORKSTATION)00Success: CPUS_LOGON shown, Winlogon converts the logon into an unlock of the existing session (logon type 7)absent1Failure, 0xC000006D / 0xC00000E5, no LSA authentication ## 7. Impact

  - Users in domains that enable "Hide entry points for Fast User Switching" can intermittently fail to unlock with correct credentials after a background Group Policy refresh.

  
     - The message suggests a wrong password, so users retry repeatedly or restart the PC.

     
        - Credential providers have no supported way to ask LogonUI to restart the usage scenario. `ICredentialProviderEvents::CredentialsChanged` re-enumerates credentials but does not call `SetUsageScenario` again, so the condition cannot be recovered from a provider.
```   ## 8. Workaround

   Dismiss the error dialog (OK), then do not interact with the sign-in screen for about 30 seconds. When the lock screen curtain returns, sign in again. Leaving the error dialog open does not recover.
   
   ## 9. Questions to Microsoft

1. Is it expected that the Registry CSE removes and rewrites `Policies\System` values non-atomically during refresh, even when the resulting value does not change?
   
   1. Can Winlogon/LogonUI decide the unlock scenario independently of a transient missing `HideFastUserSwitching` value, or re-evaluate the scenario when the submission fails with STATUS_INTERNAL_ERROR?
   
   1. Is there a supported setting to avoid the gap (for example "Configure registry policy processing" options) without side effects on other registry-based policies?
   
Windows for business | Windows Server | Directory services | Deploy group policy objects
0 comments No comments

2 answers

Sort by: Oldest
  1. Daphne Huynh (WICLOUD CORPORATION) 2,620 Reputation points Microsoft External Staff Moderator
    2026-10-28T11:01:18.08+00:00

    Welcome to Microsoft Q&A!

    Thank you for taking the time to provide such a detailed reproduction, including the timing measurements, event-log observations, and the clean test using the built-in Password Credential Provider without a third-party credential provider.

    Based on the evidence you collected, this does not appear to be a normal incorrect-password or Active Directory authentication failure.

    The most important observation is that the failing credential submission does not produce the expected 4624, 4625, or 4648 authentication events, while the same credentials succeed after LogonUI returns to the normal workstation-unlock state. Your clean reproduction also makes a third-party credential provider much less likely to be responsible for the behavior.

    • Explanation

    Windows credential providers operate according to a usage scenario supplied by LogonUI through ICredentialProvider::SetUsageScenario.

    In this case, the relevant scenarios are CPUS_LOGON and CPUS_UNLOCK_WORKSTATION. CPUS_UNLOCK_WORKSTATION represents the workstation-unlock scenario, while CPUS_LOGON is associated with the logon/unlock sign-in experience.

    The HideFastUserSwitching policy is a computer-level policy, so its state applies to the device's sign-in experience rather than to an individual user.

    Your trace shows a very specific sequence:

    1. HideFastUserSwitching is temporarily absent while Group Policy registry processing is occurring.
    2. The workstation is locked during that interval.
    3. LogonUI initializes and the credential provider receives CPUS_LOGON.
    4. HideFastUserSwitching is subsequently restored to 1.
    5. The existing LogonUI instance remains in the previously initialized state.
    6. Submitting the correct password then returns 0xC000006D (STATUS_LOGON_FAILURE) with 0xC00000E5 (STATUS_INTERNAL_ERROR).
    7. No corresponding LSA authentication event is recorded for that failed submission.
    8. After the sign-in UI returns to the lock-screen curtain and the credential providers are enumerated again under the unlock scenario, the same credentials successfully unlock the existing session and generate logon type 7.

    Taken together, these observations are consistent with a timing-sensitive LogonUI/Winlogon state issue occurring around Group Policy processing, rather than evidence that the user's password is actually invalid.

    However, there is an important distinction here: this is an interpretation of the behavior demonstrated by your traces. It should not be treated as confirmation that Microsoft has documented this as a Windows 11 24H2 known issue or confirmed product defect.

    Unfortunately, I could not find public Microsoft documentation confirming that the temporary removal of this particular registry value during Registry CSE processing, followed by this STATUS_LOGON_FAILURE / STATUS_INTERNAL_ERROR sequence, is documented or expected behavior.

    • Workaround

    For the condition you reproduced, I recommend using the recovery procedure that your testing has already demonstrated to work:

    Dismiss the "The password is incorrect" dialog, stop interacting with the sign-in screen, wait until the lock-screen curtain returns, and then unlock the workstation again.

    Do not continuously retry the password or leave the error dialog displayed.

    Based on your reproduction, retrying while the same LogonUI state remains active continues to fail. Once the lock-screen curtain returns, the credential providers are enumerated again and the subsequent authentication proceeds as a normal workstation unlock.

    I would not recommend changing Registry Group Policy processing options specifically to solve this issue because I could not find Microsoft documentation establishing such a configuration change as a supported workaround for this condition.

    • Next validation step

    After reproducing the issue, dismiss the error dialog and leave the sign-in screen idle until the lock-screen curtain returns.

    Then unlock the workstation with the same credentials and confirm that the successful attempt generates Security event 4624 with logon type 7.

    That would provide a useful validation that the failure is associated with the transient sign-in state rather than with the validity of the user's credentials.

    • Important notes

    I understand that the workaround alone may not be a satisfactory resolution, especially when this can occur repeatedly in an enterprise environment. The evidence you collected is detailed enough that it is reasonable to want to know whether this is expected Windows behavior and, if not, whether Microsoft can correct it.

    However, there is an important limitation to what can be established through Microsoft Q&A.

    Microsoft Q&A is a public technical community forum. It is suitable for discussing documented Windows behavior, reviewing configurations, interpreting logs and published documentation, and suggesting documented troubleshooting steps or workarounds.

    This particular case goes beyond those boundaries because the remaining questions concern the internal interaction between the Group Policy Registry client-side extension, LogonUI, Winlogon, credential-provider usage-scenario selection, and authentication processing.

    Specifically, answering the following questions definitively would require product-level investigation:

    • whether temporarily removing and recreating HideFastUserSwitching during Registry CSE processing is expected implementation behavior.
    • whether LogonUI/Winlogon should independently recognize the existing locked console session when the registry value is temporarily unavailable.
    • whether Winlogon should reconsider the current credential-provider usage scenario after receiving STATUS_INTERNAL_ERROR.
    • whether the Registry policy update and LogonUI initialization are experiencing an unintended race condition
    • whether Windows should be changed to prevent or recover automatically from this state.

    Those questions cannot be answered reliably by examining public documentation alone. A forum reply also cannot determine the internal Windows execution path from source code, perform private debugging of Winlogon/LogonUI, confirm that an observed implementation detail is "by design," or classify the behavior as a Windows product defect requiring a servicing change.

    For that reason, I do not want to label this definitively as a Windows 11 24H2 bug based only on the reproduction, even though the evidence strongly isolates the condition. Doing so would go beyond what can be supported by the currently available documentation.

    If you require a root-cause determination or permanent product-level resolution, the appropriate next step would be to open a Microsoft Support request.

    Your existing data should be particularly useful for that investigation because you already have:

    • a deterministic reproduction.
    • a reproduction without a third-party credential provider.
    • timestamps showing the registry-value transition.
    • the observed CPUS_LOGON versus CPUS_UNLOCK_WORKSTATION behavior.
    • the 0xC000006D / 0xC00000E5 result.
    • confirmation that no corresponding LSA authentication events occur during the failed submission.
    • confirmation that the same credentials subsequently succeed as logon type 7.

    That level of evidence should allow the issue to be investigated beyond what can reasonably be established in a public forum.

    I realize this does not provide the definitive explanation or permanent fix you were hoping for. However, I would rather be transparent about the limits of what can be confirmed from published documentation than present an interpretation of internal Windows behavior as an official Microsoft product conclusion.

    References:

    CREDENTIAL_PROVIDER_USAGE_SCENARIO (credentialprovider.h) - Win32 apps | Microsoft Learn

    nuget.exe Credential Providers | Microsoft Learn

    CredentialProviders Policy CSP | Microsoft Learn

    Configure Windows Hello for Business | Microsoft Learn

    If you find it useful, please click Accept Answer.

    Thank you for choosing Microsoft Q&A.

    Was this answer helpful?

    0 comments No comments

  2. Deleted

    This answer has been deleted due to a violation of our Code of Conduct. The answer was manually reported or identified through automated detection before action was taken. Please refer to our Code of Conduct for more information.

    1 deleted comment

    Comments have been turned off. Learn more

Your answer

Answers can be marked as 'Accepted' by the question author and 'Recommended' by moderators, which helps users know the answer solved the author's problem.