OR type logical switch mapped to real switch

The total evaluations required to get to the point of executing the action remains the same, so conditionals evaluated = triggers-from-to + conditions, split either way.
But I’ll rely on your wisdom re. pre-compute, caching, logging, etc. i.e. the other parts of the ecosystem that may or may not get invoked beyond conditional logic.

You overlooked to mention that your original version will perform the evaluations more often because it triggers for any state-change. In addition, it will generate a trace even when the action isn’t executed.

EDIT

Have you tried the example I suggested?

Compare the number of traces it produces to your version. Simple test is to just change a climate entity’s operating mode.