Great question on Web Workers - yes, I'm using them! The simulation engine has three modes:

1. *Web Worker* (default for large files): Runs threshold evaluation off-thread. UI stays completely responsive. This is what kicks in for your scenario with millions of rows.

2. *Chunked processing* (fallback): Processes data in batches with setTimeout between chunks if Workers aren't available. Slower but still keeps UI alive.

3. *Synchronous* (small files only): Direct processing for datasets where the overhead of spawning a worker isn't worth it.

The threshold logic itself is pretty straightforward - for each data point, check value against warn/crit/emrg thresholds based on the operator (>, <, >=, etc.). The Worker handles iterating through potentially millions of rows and building the alert timeline.

Progress updates come back from the Worker every ~1000 rows so users see the progress bar move in real-time.