schemeshard: fix SyncTenant handling for dropped extsubdomains (#22360)

Author: ijon

ydb/core/tx/schemeshard/schemeshard_impl.cpp

@@ -5773,6 +5773,17 @@ void TSchemeShard::Handle(TEvSchemeShard::TEvSyncTenantSchemeShard::TPtr& ev, co
  return;
  }
 
+ if (PathsById.at(pathId)->Dropped()) {
+ // This could happen when root schemeshard reboots just after marking subdomain's path as dropped
+ // but before being able to begin subdomain cleanup. Then, if tenant schemeshard tablet is still alive,
+ // it will detect disconnect error in pipe-to-parent, re-establish connection and send TEvSyncTenantSchemeShard.
+ // Root schemeshard should ignore it and should not register dropped subdomain in subdomain links again.
+ LOG_WARN_S(ctx, NKikimrServices::FLAT_TX_SCHEMESHARD, "Handle TEvSyncTenantSchemeShard, at schemeshard: " << TabletID()
+ << ", ignore spurious message from dropped subdomain's schemeshard (pre cleanup)" << pathId
+ );
+ return;
+ }
+
  if (SubDomainsLinks.Sync(ev, ctx)) {
  Execute(CreateTxSyncTenant(pathId), ctx);
  }