Struct ReconnectCoordinator 

pub struct ReconnectCoordinator { /* private fields */ }

Coordinates subscription reconnects across a parallel batch sharing one bandwidth-constrained link.

A dropped subscription competing with its still-streaming siblings just starves — the handshake can’t get through a saturated pipe. So once a session has burned its fast ungated attempts, it stops fighting and parks until no sibling is streaming (the connected sessions have finished and freed their bandwidth), then reconnects into the quiet link and resumes via replayFromSlot. Stepping aside also speeds the streaming siblings up, so the batch drains and the parked sessions recover — serially, one at a time, since at a tight rate only one stream fits.

Implementations

impl ReconnectCoordinator

pub fn new() -> Self

pub fn enter(self: &Arc<Self>) -> StreamingGuard

Mark this subscription as streaming until the returned guard drops. Call once a (re)subscribe succeeds; the guard’s Drop decrements the count and wakes any sibling parked in Self::reconnect_slot. RAII so a panicking or aborted task can’t leak the count and wedge the batch.

pub async fn reconnect_slot( &self, cancel: &CancellationToken, ) -> Option<OwnedSemaphorePermit>

Park until no sibling is streaming, then take the single reconnect slot. The returned permit must be held across connect + subscribe and dropped once Self::enter has been called (or the attempt failed), so the next parked session proceeds only after this one is streaming again. Returns None if cancelled.

Trait Implementations

impl Default for ReconnectCoordinator

fn default() -> Self

Returns the “default value” for a type.