Task is ahead/behind of ShardState

[ckhan]

We use mapreduce and pipelines for various administrative jobs on GAE.
Almost every day, I'll see warnings like this in the log:

LOG 2 1468861523504730 Task 1571100778565B9A65B43-5-0 is behind ShardState 1. Dropping task.
..
LOG 2 1468861525573810 Shard 1571100778626ECE727ED-7 is not active. Possible spurious task execution. Dropping this task.
..
LOG 2 1468861525628980 ShardState is {active:False,last_work_item:Key('FileEntity', 5711710446944256, namespace='5666083260334080'),result_status:success,slice_id:3,update_time:2016-07-18 17:04:00.927030,}
..
LOG 2 1468618964079700 Task 15711250257171C594209-3-2 is ahead of ShardState 1. Waiting for it to catch up.

They seem to be innocuous (just retries?), but they are making a lot of unwanted noise in the logs.
Questions:

• what do these errors mean?
• if it's innocuous, should it be logged as warning? should we drop logging level in our fork?
• if not, what can we do to mitigate the underlying issue?

Thanks for any pointers!
-ck