Category: Online
Title: Intermittent "unable to download resource" — preloader stalls permanently, PsychoPy 2026.1.3
PsychoPy version: 2026.1.3
Browser: Chrome
OS: Windows
Running: locally via pilot mode (localhost). Study is intended to run online on Pavlovia.
Problem
The resource preloader intermittently stalls during loading and never recovers. The download counter stops advancing (e.g. “downloading 47.5 / 72”) and the experiment never starts.
The page itself remains responsive — the dialog is still clickable and the tab closes normally — but loading makes no further progress no matter how long it is left.
Refreshing does not resolve it. The refreshed page stalls at the same point, or produces this error:
* when processing a loading error
* unable to download resource: ats_g1_practice_taskB.csv (ats_g1_practice_taskB.csv)
There is no way to recover from within the experiment once it stalls.
Reproduces in two unrelated experiments
It occurs in my own experiment (72 resources: 48 JPEGs + 24 CSVs) and in a completely separate experiment written by a different member of our lab (54 resources), which failed identically:
* when processing a loading error
* unable to download resource: FaceCollection/060_o_f_n_a.jpg
Different code, different developer, different file structure — same failure.
The failing file varies between runs. It is not a specific corrupt or missing file; every file loads successfully on other runs.
Network tab evidence
When stalled, the Network panel shows one request permanently pending, 0.0 kB transferred, initiated by preloadjs.min.js, with no timeout and no retry:
167_y_m_h_b.jpg (pending) xhr preloadjs.min.js:12 0.0 kB Pending
Summary at that moment: 32 requests | 26.6 MB transferred | 53.2 MB resources
The console shows nothing beyond normal startup messages — no resource errors are logged when it hangs silently.
What we have ruled out
- Not a specific file — the failing file differs each time, and all files are verified present
- Not total payload size — occurs both with ~82 MB of images and with ~6 MB after downscaling
- Not our experiment’s code — reproduces in an unrelated experiment
- Not one machine — reproduces on two separate machines
- Not browser cache — occurs in fresh sessions
Questions
- Is this a known issue in 2026.1.3?
- Can a timeout or retry be configured on the preloader, so a hung request doesn’t block indefinitely?
- Is there a recommended approach for experiments with a large number of image resources?
This is a concern for us because the study will run online with child participants, where a permanently stalled loading screen would mean losing that participant.