Feature Request: clip time delta outliers in relative mode

Relative mode and wall clock mode are both useful for comparing absolute training time between different variants of a model where the training steps may take different lengths of time. The main advantage of relative mode is that it can cope with the situation where training doesn't start at the same time for different models. But it still doesn't do a good job with the situation where one model stops training for a while and then is restarted (either because the GPU was needed for something else or because the training crashed for some reason and had to be manually restarted). It would be useful if relative mode could detect unusually long intervals between logging to tensorboard and clip them to a reasonable value so as to be useful even when training is interrupted like this.

The issue names no files, tests, or entry points. Start by locating the relative-mode time-delta handling and determine how interrupted TensorBoard logging is represented; confirm the intended clipping threshold with maintainers. Done means relative mode remains useful after unusually long logging gaps without changing normal intervals, with coverage for interrupted training.