Test sync::syncs_header_only_forks is flaky in sc-network-test

[dmitry-markin]

With changes from paritytech/substrate#14337, test syncs_header_only_forks is flaky on a multi-threaded runtime. This is only an issue on some machines where it fails ~ every 4 hours of continuous running with other sync tests.

For some reason, the test is not flaky if run under single-threaded #[tokio::test] runtime. So, paritytech/substrate#14337 converts it into single-threaded. Nevertheless, this needs some attention, so marked as TODO in the code.

See paritytech/substrate#14337 (comment) for additional details.

[ggwpez]

Is it using thread-local state somewhere?

[dmitry-markin]

Is it using thread-local state somewhere?

That's highly unlikely.