Mongoz is a replacement for MongoDB sharding server (mongos) aimed at higher availability in less-than-absolutely-reliable network environments at expense of somewhat relaxed consistency guarantees.
Some distinguishing features of mongoz include:
Hard timeouts: each operation has to finish in predefined time. In case of network failures, clients get their errors early (i.e. they don't have to wait for i/o timeout from operating system) and thus are much less likely to face a DoS.
Early request retransmission: if there's more than one replica capable of handling a request, mongoz can be configured to retransmit a request to another replica before a timeout occurs and return whichever reply comes first. That way, even a network failure in the middle of a performing request does not neccessarily lead to any errors reported to clients.
Config caching: mongoz does not query its config servers upon a specific event (like incoming connection or authentication attempt). Instead, it keeps its local copy of the whole cluster config and synchronizes it periodically with config servers. As a result, unavailable config servers (even all of them) still don't render the whole cluster unusable.
Mongoz strictly obeys MongoDB wire protocol and thus can be used as a drop-in replacement for mongos. Moreover, it can freely coexist with mongos within a same MongoDB cluster if neccessary.
More detailed description of mongoz features can be found in its man page.