Three tests are regularly failing on macOS M1

I'm wondering if I'm the only one with these flaky tests: test_doc_processor_simple_vrl, test_indexer_on_timeout and test_indexer_partitioning. I never see them on the CI or coverage.

---- actors::doc_processor::tests::test_doc_processor_simple_vrl stdout ----
thread 'actors::doc_processor::tests::test_doc_processor_simple_vrl' panicked at 'assertion failed: `(left == right)`
  left: `DocProcessorCounters { index_id: "my-index", source_id: "my-source", num_parse_errors: 0, num_transform_errors: 0, num_docs_with_missing_fields: 0, num_valid_docs: 0, overall_num_bytes: 0 }`,
 right: `DocProcessorCounters { index_id: "my-index", source_id: "my-source", num_parse_errors: 1, num_transform_errors: 0, num_docs_with_missing_fields: 1, num_valid_docs: 2, overall_num_bytes: 397 }`', quickwit-indexing/src/actors/doc_processor.rs:671:9
note: run with `RUST_BACKTRACE=1` environment variable to display a backtrace

---- actors::indexer::tests::test_indexer_on_timeout stdout ----
thread 'actors::indexer::tests::test_indexer_on_timeout' panicked at 'assertion failed: `(left == right)`
  left: `IndexerCounters { num_splits_emitted: 0, num_split_batches_emitted: 0, num_docs_in_workbench: 0 }`,
 right: `IndexerCounters { num_splits_emitted: 0, num_split_batches_emitted: 0, num_docs_in_workbench: 1 }`', quickwit-indexing/src/actors/indexer.rs:855:9

---- actors::indexer::tests::test_indexer_partitioning stdout ----
thread 'actors::indexer::tests::test_indexer_partitioning' panicked at 'assertion failed: `(left == right)`
  left: `IndexerCounters { num_splits_emitted: 0, num_split_batches_emitted: 0, num_docs_in_workbench: 0 }`,
 right: `IndexerCounters { num_splits_emitted: 0, num_split_batches_emitted: 0, num_docs_in_workbench: 2 }`', quickwit-indexing/src/actors/indexer.rs:1033:9

Start by reproducing the three named tests on macOS M1 and inspect the assertions in quickwit-indexing/src/actors/doc_processor.rs, quickwit-indexing/src/actors/indexer.rs, and their surrounding test setup. Compare the observed counters with the expected values and determine why these tests behave differently on that platform. Done means the tests no longer fail intermittently on macOS M1 while retaining their intended assertions.