[cluster2600]

Fixes #6053

Problem

The # Disconnection section of the mpsc module docs stated:

As such, Receiver::poll returns Ok(Ready(None)).

This is incorrect in two ways:

1. Wrong condition: When all Sender handles are dropped, recv does NOT immediately return None. Messages already buffered in the channel can still be received. recv only returns None once all senders are dropped and the channel has been fully drained.

2. Wrong method name: Receiver::poll is not a public API method. The public methods are Receiver::recv and Receiver::poll_recv.

Fix

Update the documentation to accurately state that recv returns None (and poll_recv returns Poll::Ready(None)) only once all senders have been dropped and all buffered values have been received.

[Darksonn]

Thanks.

[cluster2600]

The test failure test_nested_block_in_place_with_block_on_between seems to be a panic in num_idle_threads underflow. This looks unrelated to this docs-only change (which only touches tokio/src/sync/mpsc/mod.rs). Is this a known flaky test or an issue with the CI environment?

[Darksonn]

Yes it was introduced recently and we're working on fixing the flaky test.