Rule for light switching - "predicted to become"

[rossko57]

The prediction comes from the autoupdate feature, probably in collusion with whatever binding is involved here.
Details, see

Autoupdate - a primer

I’ve found myself having to investigate and/or explain some of the features of autoupdate over time, and thought an in-depth view would be helpful. These notes apply to openHAB version 2.5 Overview Autoupdate is primarily a “speed up” feature for openHAB Items. Mostly we can ignore it, but sometimes you may get an unwanted or unexpected effect, or you may wish to exercise more control. To fully understand autoupdate, we must understand the difference between command and state of Items. If w…

I think the area of interest is likely to be

“Linked Item with broken channel or an OFFLINE Thing, or binding veto”

which I think is about the only circumstance where a prediction overrides a command. The binding has said “nah, can’t do that, somethings offline”

So your real problem is -

@Torsten555

Sometimes the rule does not work as expected

Sometimes the command cannot be implemented - connectivity, busy with something else?
There are/were a few zigbee things that applied a veto in the binding, which would look like this, but that’s permanent not occasionally.

Looking at timestamps, 26mS seems a long time between command and prediction. Is this a heavily loaded system? Is it perhaps busy doing some kind of error recovery?