This paper discusses some dependability problems (e.g. message delays) related to clock drifts in the processing nodes of real-time networks based on event and time triggered interfaces. We show that having detected the clock drift, it is reasonable to delay the shutdown of the network node whose clock synchronization algorithm is not able to track a high clock drift. We observed that transient clock drifts, which last up to one second and are as high as 2e-2 sec/sec do not increase message delays above safety-critical required levels. We verified this by injecting transient faults into our new clock model and measured the message delay between a pair of network nodes. The performed experiments revealed that time triggered interfaces are more sensitive to transient clock drift disturbances than event triggered interfaces.