Release candidate 4 for 3.6.0. Two bug fixes caught during the rc3 bake — both visible regressions from earlier rc work. No new features.
Fixed in rc4
Static-frame mode hid the radar layer on first pan
When past_minutes: 0 (one frame, no animation), the radar player's _initRadar showed the static preview but never called _startLoop (which requires ≥ 2 loaded slots). The _prev1Slot tracker stayed at its initial -1. On the first pan or zoom, onNavPaused → _stopLoop → _settleVisibility looped through all loaded radar layers and set opacity: 0 on every slot whose index didn't match -1 (i.e., all of them, including the only visible one). The radar precipitation overlay vanished until a full re-init was forced.
Fixed by initialising _prev1Slot to the visible-slot index when the static preview goes up. Animated mode unaffected — its _startLoop overwrites the value one iteration later.
Wind streamlines drew giant line segments after tab returns from hidden
Browsers throttle requestAnimationFrame to ~1 Hz on background tabs (or pause it entirely). When the tab became visible again, the wind overlay's motion-compensation math (motionScale = elapsedMs / 33.33) saw a multi-second gap and scaled per-frame motion by hundreds, producing massive single-frame jumps that drew very long streak segments across the canvas.
Fixed by capping the motion-comp dt at 2× the target frame interval (~133 ms / motionScale max 4) so any single frame can move at most 4× the per-frame distance (vs. ~2× normal at 15 fps). Particles "lose" the hidden time but resume at a sensible pace on visibility return.
Cumulative since 3.5.0
Same as rc3 plus the two fixes above. See v3.6.0-rc1 notes for the full 3.6 feature inventory.
Upgrade notes
No breaking changes. No new config keys. Pure bug fixes. If nothing surfaces during the rc4 bake, this is what 3.6.0 ships as.
Full Changelog: v3.6.0-rc3...v3.6.0-rc4