You should check this old issue in the Laravel framework.
Quoting Taylor Otwell,
The issue is in the query that migrates jobs that have been reserved too long to being available. We have a bunch of workers trying to execute that query simultaneously. We could simply ignore problems with that query and assume another worker is already doing it.
The way I mitigated the issue is use Redis to handle the queue. And of course, this is the reason why we use Laravel Horizon.
And based on the issue thread above, this problem should be fixed in the Laravel 5.2.
Hope this helps.