Try to fix dump file upload

Author: vmoroz

.ado/jobs/e2e-test.yml

@@ -184,6 +184,17 @@ jobs:
                 displayName: Disarm crash-simulation sentinel (TEST ONLY)
                 condition: always()
 
+            # Always disarm the hang-simulation env var so the post-failure
+            # `Update snapshots` step (which also runs `yarn e2etest`) does not
+            # re-trigger the hang and burn 10 minutes of dead time. Setting an
+            # ADO variable to empty string clears it for subsequent steps.
+            - ${{ if eq(parameters.simulateHangForTesting, true) }}:
+              - pwsh: |
+                  Write-Host "##vso[task.setvariable variable=RNW_SIMULATE_HANG]"
+                  Write-Host "Hang simulation disarmed (RNW_SIMULATE_HANG cleared)"
+                displayName: Disarm hang-simulation env var (TEST ONLY)
+                condition: always()
+
             # On test failure, snapshot any lingering RNTesterApp-Fabric / node
             # processes before subsequent steps (or the agent) tear them down.
             # WER LocalDumps only fires on actual crashes; this catches hangs
@@ -202,8 +213,12 @@ jobs:
                     $dumpPath = Join-Path "$(CrashDumpRootPath)" ("hang_{0}_{1}.dmp" -f $name, $_.Id)
                     Write-Host "Capturing full dump of $name (pid $($_.Id)) to $dumpPath"
                     & $procDump -accepteula -ma $_.Id $dumpPath
+                    Write-Host ("ProcDump exit code: {0} (non-zero is normal — encodes the dump count written)" -f $LASTEXITCODE)
                   }
                 }
+                # ProcDump uses non-zero exit codes to encode the number of dumps written.
+                # Force a clean PowerShell exit so the step doesn't show as a warning.
+                exit 0
               displayName: Capture dumps of surviving test processes
               condition: and(failed(), eq(variables.StartedFabricTests, 'true'))
               continueOnError: true