OSWeb runtime broken on Mindprobe — _setupClickScreen null error

Hello,

I'm experiencing a complete outage on jatos.mindprobe.eu where **all** my OSWeb studies — including previously deployed and confirmed-working ones — fail on the "Building screen" stage. This appears to be a recent server-side regression in the OSWeb runtime.

### The error

Browser console output:

  PixiJS 5.3.12 - WebGL 2

  text.es.js:1307 Canvas2D: Multiple readback operations using getImageData are faster...

  transfer.js:121 file pool embedded in HTML (N files)

  transfer.js:246 All audio files have been loaded and decoded

  screen.js:202  Uncaught (in promise) TypeError: Cannot read properties of null (reading 'length')

            at Screen._setupClickScreen (screen.js:202)

            at Parameters._initialize (parameters.js:24)

            at Runner._setupContext  (runner.js:143)

Tracing the call chain: `this._runner._welcomeText` is `null` when passed to `_setupClickScreen`. The `typeof text === "undefined"` guard at the top of the function does not catch `null` (because `typeof null === "object"`), so the function continues and crashes when reading `.length` on the null text.

### Evidence this is server-side, not study content

1. **Previously-working deployed studies** that ran fine for weeks now produce the same error without any changes from my side.

2. **A reference .osexp** I successfully ran earlier this year now fails identically when re-uploaded.

3. **Multiple .osexp variants** with different headers (with/without `set background`, with/without `set jatos_uuid`, Python `True/False` vs legacy `yes/no` booleans, ASCII vs Turkish titles, with/without inline_javascript blocks, with/without `run_if` conditions) all produce the identical error.

4. Tested across **two Mindprobe accounts** and multiple browsers (Chrome, Firefox, incognito).

### Suspected cause

This looks like a recent OSWeb runtime update or transpilation pipeline change on the Mindprobe server. The failure pattern (identical error on unchanged previously-working studies) is hard to explain otherwise.

Thanks,

Büşra