Hi Daiwen,
Please feel free to send the experiment link to support@pcibex.net if you feel more comfortable sharing it privately
The random blank image issue might have difference sources. One is, you might have different groups in your CSV file and only some rows have invalid references, resulting in the error showing only when you run the experiment in that group—if you have an early SetCounter command, simply refreshing the page could be enough to switch group. Another explanation would be that the server hosting the images refuses to serve some of them in certain conditions (eg. repeated requests over a short time window)
The error message “ERROR: There must be some items in the running order!” appears when there is a syntax error in your script that prevents it from creating the items variable (or when you simply don’t create any items, but you said that’s not your case). Simply refreshing a functional experiment should not make the error message appear, unless you edit the script in the meantime to introduce syntax errors. One explanation I could imagine is the farm maintains multiple parallel copies of your script and randomly serves one that does contain a syntax error. Maybe try to save a backup of your script (that you know to be functional), delete the script file from your project and recreate it using the backup, and see whether the problem persists
Jeremy