unhandled_panic(UnhandledPanic::ShutdownRuntime) not working when endless interval loop in tasks

[markus2330]

Version

Probably only tokio_unstable v1.35.0 relevant for this issue but just in case:
cargo tree:

├── async-ssh2-tokio v0.7.1
│   │   │   ├── tokio v1.35.0
│   │   │   │   └── tokio-macros v2.2.0 (proc-macro)
│   │   │   ├── tokio-stream v0.1.14
│   │   │   │   └── tokio v1.35.0 (*)
│   │   ├── tokio v1.35.0 (*)
│   │   └── tokio-util v0.7.2
│   │       ├── tokio v1.35.0 (*)
│   └── tokio v1.35.0 (*)
│   │   ├── tokio v1.35.0 (*)
│   │   ├── tokio-util v0.7.2 (*)
│   │   ├── tokio v1.35.0 (*)
│   │   ├── tokio v1.35.0 (*)
│   │   └── tokio-native-tls v0.3.1
│   │       └── tokio v1.35.0 (*)
│   ├── tokio v1.35.0 (*)
│   ├── tokio-native-tls v0.3.1 (*)
├── signal-hook-tokio v0.3.1
│   └── tokio v1.35.0 (*)
├── tokio v1.35.0 (*)
└── tokio-util v0.6.10
    └── tokio v1.35.0 (*)

.cargo/config.toml

[build]
rustflags = ["--cfg", "tokio_unstable"]

Platform
The output of uname -a (UNIX):

Linux markusbyte 5.10.0-26-amd64 #1 SMP Debian 5.10.197-1 (2023-09-29) x86_64 GNU/Linux

Description

I tried this code:

use futures::future::join_all;
use std::panic;
use std::time::Duration;
use tokio::runtime::UnhandledPanic;
use tokio::spawn;
use tokio::time::interval;

fn main() {
	tokio::runtime::Builder::new_multi_thread()
		.unhandled_panic(UnhandledPanic::ShutdownRuntime)
		.enable_all()
		.build()
		.unwrap()
		.block_on(async {
			let mut tasks = vec![];
			tasks.push(spawn(async {
				let mut interval = interval(Duration::from_millis(10));
				loop {
					interval.tick().await;
				}
			}));
			tasks.push(spawn(async {
				panic!("boom");
			}));
			join_all(tasks).await;
		});
}

I expected to see this happen:

thread 'tokio-runtime-worker' panicked at src/main.rs:23:17:
boom
note: run with `RUST_BACKTRACE=1` environment variable to display a backtrace
exit code 101

Instead, this happened:

thread 'tokio-runtime-worker' panicked at src/main.rs:23:17:
boom
note: run with `RUST_BACKTRACE=1` environment variable to display a backtrace

The process is still running and does not quit.

[maminrayej]

This is intended behavior. From the original issue:

When the multi-threaded runtime is selected, these option would have no effect.
Implementing for the multi-threaded runtime would be required before stabilizing
the API but because the implementation is much harder, we should first gather data.