storage_controller: rename failpoint and make it pausable

The same failpoint is used for a new test by a follow up commit
and that needs a pausable failpoint.

storage_controller/src/reconciler.rs

@@ -14,7 +14,6 @@ use std::sync::Arc;
 use std::time::{Duration, Instant};
 use tokio_util::sync::CancellationToken;
 use utils::backoff::exponential_backoff;
-use utils::failpoint_support;
 use utils::generation::Generation;
 use utils::id::{NodeId, TimelineId};
 use utils::lsn::Lsn;
@@ -824,7 +823,7 @@ impl Reconciler {
                 .handle_detach(self.tenant_shard_id, self.shard.stripe_size);
         }
 
-        failpoint_support::sleep_millis_async!("sleep-on-reconcile-epilogue");
+        pausable_failpoint!("reconciler-epilogue");
 
         Ok(())
     }

test_runner/regress/test_storage_controller.py

@@ -2406,7 +2406,14 @@ def test_storage_controller_step_down(neon_env_builder: NeonEnvBuilder):
     env.storage_controller.tenant_create(tid)
 
     env.storage_controller.reconcile_until_idle()
-    env.storage_controller.configure_failpoints(("sleep-on-reconcile-epilogue", "return(10000)"))
+    env.storage_controller.configure_failpoints(("reconciler-epilogue", "pause"))
+
+    def unpause_failpoint():
+        time.sleep(2)
+        env.storage_controller.configure_failpoints(("reconciler-epilogue", "off"))
+
+    thread = threading.Thread(target=unpause_failpoint)
+    thread.start()
 
     # Make a change to the tenant config to trigger a slow reconcile
     virtual_ps_http = PageserverHttpClient(env.storage_controller_port, lambda: True)
@@ -2421,6 +2428,8 @@ def test_storage_controller_step_down(neon_env_builder: NeonEnvBuilder):
     observed_state = env.storage_controller.step_down()
     log.info(f"Storage controller stepped down with {observed_state=}")
 
+    thread.join()
+
     # Validate that we waited for the slow reconcile to complete
     # and updated the observed state in the storcon before stepping down.
     node_id = str(env.pageserver.id)