Interestingly i have just tested it out and it seems that there is some kind of a rounding thing coming up with values being exactly the threshold value. If value changes next fraction of the threshold then this does not happen.
A quick test
Create a virtual point with exact settings
Current value 1
Set point to 1.01
No logging
Set point back to 1.0
Set point to 1.011
Logging appears
If going to other way around
Current value 1.01
Set point to 1.00
Logging appears
I am guessing datalogging does not happen if new value is greater than current value in the defined threshold but if new value is lower then current value in the defined threshold then logging works correctly.
I might also fish out an idea that @Nurr polling time is related to the issue (I guess it might be once per minute?) where the cached value is the new over threshold limit but it does not get logged but still triggers a reset for interval logging and then it has been the same for 5 minutes and then it gets logged by the interval logging sequence.
But still even though this happens, interval logging still works in my instance so I guess this information is not actually so relevant.