Well a system that aims for security can not rely on self-asserted time, and there is no distributed secure time.  If you start from a false assumption people will systematically abuse your assumption.

This is why for optimistic parallel discrete event simulation Lamport clocks were invented (some kind of vector of local clocks based on observed messages, to partially order transactions, and detect out of order relayed messages, without relying on clock sync, the only assumption is the clocks monotonically increase, they dont even have to be remotely accurate to wall-clock) - you want to progress the simulation with optimistically and roll-back when you later discover causality violations.  But rollback is not such a hot idea in bitcoin, then your n-confirmed coins disappear at uncontrollable wall-clock delays.  You do not ideally want to rely on clock accuracy or secure clock convergence protocol in a network with symbil attacks.  Bitcoin may already depend slightly too much on clock convergence.

Also in crypto protocols for freshness proofs (anti-replay) people use a relying party generated nonce rather than time.

Heh maybe you can find a new 8th use for mining - a loose (+/- 10mins)  sybil resistant clock convergence (same mining argument against byzantine generals).  Maybe thats already happening I never did read how bitcoins internal clock sanity rules work.

Adam