It's possible for a task to get lost if there's a communications glitch between the host and the server.
We used to have the server configured to resend such lost tasks. Unfortunately, enabling that option creates a worse problem where tasks get stuck in limbo forever. It turns out that it's better to leave it turned off. This way the last tasks simply time out.
It's impossible to check whether this is actually what happened in this instance, but it's a possibility.
My lucky number is 75898524288+1