Hi Daoxin,
Discrepancies between the two links often have to do with them running different groups: after taking the demonstration link a few times, you might be running in, say, group 2, but if you’ve never completed the study with the data-collection link before, you’ll be running it in the first group using that link. If there are problematic references in your table(s) for the first group, but not the second, then you’ll only experience issues with the data-collection link (although you can force a group when running your experiment with either link using server.py?withsquare=N)
So first thing to do is double-check your file references: make sure each filename from your tables is found across the zip files and the files in your project’s Resources folder
That being said, I couldn’t really replicate your specific issue, because the two high* zip files you preload are several hundreds MBs each, and my browser would fail to complete the background downloads, so some video files (as part of a zip file) simply were never downloaded and therefore could not be preloaded at all, regardless of which group or link I would use. That’s another potential source of discrepancy: maybe the zip files had already been downloaded before during your test runs with the demonstration link and you have them in your browser’s cache, but running the data-collection link might have prompted new download requests that did not complete, as was the case for me. It would be worth investigating whether/how you could make your video files, and so ultimately your zip files, lighter so that they can be properly downloaded in time
Jeremy