I noticed (by observing the impact on the latency between Adobe Audition's input and Adobe Audition's output) that Adobe Audition computes the effects that have been disabled. Is there any point in doing so?
Yes. There is a very good reason why DAW's do this. Specifically, the DAW will always take into account channel latency across all tracks currently active in the DAW whether the inserted plugins are active or not.
Imagine a multi-track project where you are tuning the sound of a particular track during playback. The required operational condition would be to ensure that the 'enabling' and 'disabling' of plugins on a particular track is seamless and doesn't affect the overall sound of the track other than to remove - albeit temporarily - the processing that the particular plugin is providing.
If the overall latency of that track were to change during that operation, it is likely to be detrimental to the overall operation as there would be a massive change in phase and a discontinuity in the audio stream, audible as a massive dropout or click.
Consequently, the standard design convention with these sorts of systems is to calculate the latency - or 'channel delay' - based on the currently inserted plugins - whether they are active or not - and maintain this channel delay consistently despite any plugins being disabled. This allows all tracks to maintain consistent phase and for the sound to be accurate and seamless even when plugins are randomly disabled or enabled during playback.