A small implementation detail made our Three.js particle playback much more predictable across phones and desktop displays: accumulate emission using delta time instead of spawning a fixed number of particles per frame.

The pattern is: add rate dt to a fractional accumulator, emit floor(accumulator), and carry the remainder forward. Clamp large dt values after tab switches, and keep a hard cap on live particles so the effect stays bounded. This prevents a 120 Hz display from emitting twice as many particles as a 60 Hz display.

We included the complete game-loop context, mobile pixel-ratio limits, object pooling notes, and a portable NixieFX runtime example here: https://nixiefx.com/threejs-game...

For teams shipping Three.js games, what has been the bigger source of variance: refresh rate, transparent overdraw, or garbage-collection spikes?