randomizer works based on it’s own counter not on number of completes. while test the survey if you will cross the randomizer and will go back so will receive different element. Or if you close the randomizer after crossing it, so again in next run you will receive different element. If you want to check the counter so check randomizer counter instead of complete runs.

even after deleting your test run, you will need to reset the counter of randomizer element.

since respondent back button will be disabled, so they will not be able to go back after crossing it and with help of unique link if respondent would close the survey in between so they will start the survey from same state where they left earlier. But if your see multiple partial responses so can get some difference in counts. but it’s rear case.

Thank you so much.

I checked the count as you said, over all its balanced. That is 8:8:8:9. Altogether 33 counts. 

In the excel sheet of data as well I got 33 responses(completed and not, altogether). 

The 10 responses with 5:2:2:1 is the last 10 responses in the excel sheet. Since the number of responses 33 matches in both, I don’t understand such a ratio difference. 

Previews and testing were all done before taking this 10 responses. So, I’m still having the doubt that they didn’t maintain this even presentation for each units, ie, 4 consecutive survey form views shows 4 different treatments.