diff --git a/tools/podium-repro/loop.ps1 b/tools/podium-repro/loop.ps1
index bf2ff87..d7e1457 100644
--- a/tools/podium-repro/loop.ps1
+++ b/tools/podium-repro/loop.ps1
@@ -28,6 +28,7 @@ param(
   # play. cdbrun.txt's PEB trick is dead on Windows 26200 (see README).
   [string]$CdbScript = 'cdbrun-gflags.txt',
   [int]$RaceSeconds = 45,
+  [string]$ReleaseExe = 'C:\VWE\RP412\Release\rpl4opt.exe',
   # Run without the heap instrument on purpose. Says so in the tally.
   [switch]$AllowNoPageHeap
 )
@@ -88,6 +89,34 @@ if (-not $pageHeapOn) {
   Log 'page heap ON (IFEO GlobalFlag has FLG_HEAP_PAGE_ALLOCS)'
 }
 
+#-----------------------------------------------------------------
+# Put THIS build in the sandboxes.
+#
+# Each pod dir keeps its own copy of the exe, planted by setup.ps1. Run a
+# batch without refreshing it and you test whatever was built whenever
+# setup.ps1 last ran - so a freshly fixed bug reproduces perfectly, on the
+# binary that still has it, and reads as the fix having failed. That
+# happened: the fix went in at 11:20 and a 20-launch batch trapped at the
+# same address on the 09:33 exe.
+#-----------------------------------------------------------------
+if (-not (Test-Path $ReleaseExe)) {
+  Log "ReleaseExe missing: $ReleaseExe - build Release first"
+  exit 1
+}
+$relHash = (Get-FileHash $ReleaseExe -Algorithm MD5).Hash
+$refreshed = 0
+for ($i = 0; $i -lt $PodCount; $i++) {
+  $podDir = Join-Path $Scratch ('pod' + [char](65 + $i))
+  if (Test-Path $podDir) {
+    Copy-Item $ReleaseExe (Join-Path $podDir 'rpl4opt.exe') -Force
+    $refreshed++
+  } else {
+    Log "pod dir missing: $podDir - run setup.ps1 -PodCount $PodCount first"
+    exit 1
+  }
+}
+Log "exe: $($relHash.Substring(0,12)) built $((Get-Item $ReleaseExe).LastWriteTime.ToString('MM-dd HH:mm:ss')) -> $refreshed pod(s)"
+
 #-----------------------------------------------------------------
 # A quiet machine before each launch.
 #