diff --git a/app/donglet/app-g031.toml b/app/donglet/app-g031.toml
index 94f077e58..3cfacc6b5 100644
--- a/app/donglet/app-g031.toml
+++ b/app/donglet/app-g031.toml
@@ -15,8 +15,9 @@ name = "task-jefe"
 priority = 0
 max-sizes = {flash = 4096, ram = 512}
 start = true
-stacksize = 368
+stacksize = 192
 notifications = ["fault", "timer"]
+features = ["no-panic", "nano"]
 
 [tasks.sys]
 name = "drv-stm32xx-sys"
diff --git a/task/jefe/Cargo.toml b/task/jefe/Cargo.toml
index 820b47b59..bae3c207f 100644
--- a/task/jefe/Cargo.toml
+++ b/task/jefe/Cargo.toml
@@ -31,6 +31,7 @@ build-util = { path = "../../build/util" }
 [features]
 dump = []
 nano = [ "ringbuf/disabled" ]
+no-panic = [ "userlib/no-panic" ]
 
 # This section is here to discourage RLS/rust-analyzer from doing test builds,
 # since test builds don't work for cross compilation.
diff --git a/task/jefe/src/main.rs b/task/jefe/src/main.rs
index 2d48acd1d..331cdd05f 100644
--- a/task/jefe/src/main.rs
+++ b/task/jefe/src/main.rs
@@ -310,9 +310,19 @@ impl idol_runtime::NotificationHandler for ServerImpl<'_> {
             let mut next_task = 1;
             while let Some(fault_index) = kipc::find_faulted_task(next_task) {
                 let fault_index = usize::from(fault_index);
-                next_task = fault_index + 1;
-
-                let status = &mut self.task_states[fault_index];
+                // This addition cannot overflow in practice, because the number
+                // of tasks in the system is very much smaller than 2**32. So we
+                // use wrapping add, because currently the compiler doesn't
+                // understand this property.
+                next_task = fault_index.wrapping_add(1);
+
+                // Safety: `fault_index` is from the kernel, and the kernel will
+                // not give us an out-of-range task index.
+                //
+                // TODO: it might be nice to fold this into a utility function
+                // in kipc or something
+                let status =
+                    unsafe { self.task_states.get_unchecked_mut(fault_index) };
 
                 // If we're aware that this task is in a fault state, don't
                 // bother making a syscall to enquire.