Good logging for me involves:

- aggregation from multiple systems

- one-line per event (so I can use unix tools)

- (ideally) enough information to replay some action in the event of a bug

- being able to live-tail it within seconds

- putting a unique, grepable identifier next to each printed output