Do you have portions of the code running in parallel?
If your code is executed in sequence, then you should not split your error cluster wire.
You can always post your code (or a screen capture .png) to show us what you mean.
If you have parallel threads of code, then yes, split the error cluster wire for each parallel thread. You then re-combine them using the Merge Error function. I think it is located within the dialog section of the Functions Palette. Sorry, I do not have LV installed on this PC, so I cannot accurately guide you to its location. You can do a search (top right of Functions Palette) for Merge Error (or simply "Error").
You should only have a single Error In at the left of the diagram and a single Error Out to the right of the block diagram.
Also, within your While Loops, you should only have 1 error cluster line. I'll edit the image to show where / what I mean.
You are using 2 optical power meters. Do you really want to read them at (nearly) the same time? If so, you may want to make sure that the sub-vi's are set to re-entrant. If that is the case, then yes, you would split the error cluster wire before and recombine after the OPM configure & read, all within he while loop.
I'll go make comments on the image.
Be back soon
PS: The code looks familiar. You obviously work in Photonics. Which country / state / province are you located? Just curious.
Here is your modified image with comments.
Doing so will already help to clean up some unwanted wires.
For an event structure loop, if you use it to respond to operator events (ie: push a button with mouse), you do not need to wire the -1 value. You also shouldn't use the 200ms delay. Each iteration will wait for an event to happen, so no need for the delay.
The top look shouldn't be to generate user events (as indicated by your comment at the bottom), but rather to handle user events. The "Start Scan" button should be one such user events. But, that would mean changing your architecture, because it looks like you want the bottom loop to run continuously..
You may not need to have a seperate loop to plot the data from queue. I can't see the code. I suspect you want to update the plot for every measurement that you take.
Can you describe the overall picture of what the application does? There may be ways to simplify the code further.
If you post your code, I'll be able to look at it in the evening.