This feature has been reviewed with the latest releases:

1.0.0.2108 – incremental changes
-- Altered Auto Extra Delay process to better manage display/probe/monitor issues
(Some probe/setting/display combinations will not work reliably with the Auto function)
-- Added Error Message for Auto Extra Delay failure due to probe/settings

1.0.0.2103 – incremental changes
-- Increased decimal places for Extra Delay value
-- Corrected potential failure with Extra Delay Auto function

The review has brought to change the previous white and near black patches to 92% yellow and cyan ones, all of them display following the settings in the Profile settings page.

I have tested that feature on my Oled, using a Prometheus, a Klein and a I1d3 as probe, pgen as pattern generator and changing the source frame frequency to 120, 60 and 24 Hz
To have stable and repeatable results it was necessary to reduce the PC load at minimum. Being the connection to the Pgen IP based, any other communication load was eliminated.
For checking if the extra delay is correct, I have used a Grey & Primary Ramp (I have used also a focused patch set (annexed) where each white level is immediately followed by the RGB ones). Looking at the RGB separation graph I could immediately detect issues with extra delay.
Here the method I have used at the end of my work.

• First decide your settings (computed delay is quite sensitive to stabilisation, both level and value). I have used a white patch with level = 141, over a full scale at 8 bit and a FS stabilisation of 250 msec.
• Then run the auto measure for 10 times and chose the max value (with my Prometheus I have gotten between 109 and 172 msec at 24 Hz, 0 and 39 at 60 Hz and all 0 at 120 Hz.
• Obviously, the lower FR adds a need of an extra delay. Keep also in mind that the basic delay is set by CS at 100 msec. These numbers show also that my panel adds also some own extra delay in displaying the patch.
• Then take the maximum value and compare the RGB separation graph you have gotten with a "secure" extra delay (I have used 500 msec)

Here an example, showing how max extra delay computation has given good results (24 Hz case).



My final notes:
The extra delay measurement gives quick info about the range of the needed extradelay. By repeating it I could get the correct value to be used.
Measurement below 100 msec are capped to 0. This is somehow a loss of info. Knowing the actual value might be useful if you want to add your own margin.