[collectd] Reliable output delivery?

[Ryan Lynch]

Collectd can support multiple output plugins, and at least some output
plugins (network) support parallel output to multiple remote hosts
(via multicast, or multiple unicasts). Very useful functionality.

My question is: Does Collectd's core support any other features
relevant to output reliability? Some possible examples that come to
mind:

 * Detecting an unavailable output destination (remote host, local
file, etc), either based on a failed write operation or some
protocol-specific event, and then flagging that destination as being
unusable.

 * Periodically re-checking the state of an unavailable destination,
and flagging it as being usable when it has returned, again.

 * Spooling failed write payloads locally in memory/disk, and then
re-trying any spooled writes when the output destination becomes
available, again.

 * Failover output destinations, which are only used when some other
(primary) output destination is unavailable.

Also: Do any other output plugins (besides network) have
reliable-delivery features?