Put a debug node showing the input to the vcgencmd node and check there is not some circumstance causing it to be called more often that you think.
Another possibility is that you are using the command somewhere else, either in another flow in node-red or outside node-red.
Well, I will admit I am using it in two places on the same machine.
I may have to optimise it so it is only used once.
However:
All things being good, I don't get outputs.
When I am playing with CSS stuff on another page (yeah: CSS is omni-present) and nothing to do with that part of the flow, I get two messages (one for each instance of the node) every 30 seconds.
This could be good for me in that if there are no spurious messages, I know the CSS is good. But I don't think that should be used as a measure.
Temporarily disable one of them (remove the input wire) and see if that fixes it. Almost certainly something you are doing with the css/deploy is accidentally synchronising them.
Just now I'm still up to my nipples in CSS buttons on another machine.
When it gets to my ankles I may look at it.
Same as the MQTT messages and just having a high level topic compared to a fuller topic and who gets the message (or spits it out) when the machine is turned on.
Yes, I agree with that, but it is/was a problem that hit me blind side.
I wanted to ask if anyone else has seen it and a simple solution.
Seems there isn't.
It would appear it needs a lot of work to resolve. Well, more than I can afford at this point in time.
So, be it I ask now or in the future - if I haven't forgotten - is a bit academic.
I do agree that it may be better. But on another point of view: I need to ask it now while I can think of it and assign priority to it according to how easy it is to fix.
Edit
I am asking because I don't know what the problem is or how easy.
If I knew it was going to be that difficult at the start, I may not have asked.
But at that point (in the past) I didn't know, so I couldn't know it was going to be that difficult.
It isn't a lot of work to disable one of the nodes and see if that fixes it. If it does then you know the cause of the problem and you can prioritise the work to fix it.