diff --git a/zebra-scan/src/storage/db.rs b/zebra-scan/src/storage/db.rs
index de5f83cbe5e..41f40f0c3b1 100644
--- a/zebra-scan/src/storage/db.rs
+++ b/zebra-scan/src/storage/db.rs
@@ -96,6 +96,8 @@ impl Storage {
     /// If a thread exited with a panic, resume that panic.
     ///
     /// This method should be called regularly, so that panics are detected as soon as possible.
+    //
+    // TODO: when we implement format changes, call this method regularly
     pub fn check_for_panics(&mut self) {
         self.db.check_for_panics()
     }
diff --git a/zebra-state/src/service/finalized_state/zebra_db.rs b/zebra-state/src/service/finalized_state/zebra_db.rs
index 155d19686aa..6229de1bca0 100644
--- a/zebra-state/src/service/finalized_state/zebra_db.rs
+++ b/zebra-state/src/service/finalized_state/zebra_db.rs
@@ -229,13 +229,16 @@ impl ZebraDb {
     ///
     /// See [`DiskDb::shutdown`] for details.
     pub fn shutdown(&mut self, force: bool) {
+        // Are we shutting down the underlying database instance?
+        let is_shutdown = force || self.db.shared_database_owners() <= 1;
+
         // # Concurrency
         //
         // The format upgrade task should be cancelled before the database is flushed or shut down.
         // This helps avoid some kinds of deadlocks.
         //
         // See also the correctness note in `DiskDb::shutdown()`.
-        if force || self.db.shared_database_owners() <= 1 {
+        if !self.debug_skip_format_upgrades && is_shutdown {
             if let Some(format_change_handle) = self.format_change_handle.as_mut() {
                 format_change_handle.force_cancel();
             }