DesecrationLantern Mire
All field notes
Field note 29Spellcasting · A playtest repair

Before the first spell.

A short preparation step moves spell resources and rendering work ahead of combat, with both wizards’ effects ready before the match starts.

Lantern Mire · Quality follow-up Prepare before combat v0.5.26
The Copper Pilgrim and three summoned creatures watch a Heavenfall lightning strike in the mire.
Actual 0.5.26 exported game at the ordinary eight-metre camera distance. Three paid summons and a real Heavenfall cast; positions, mana and target health are staged, with rival and camp drivers disabled. Full frame resized without retouching. A still image is not evidence of frame smoothness.

Shaun reported a severe pause when casting a spell for the first time. Smooth averages from a warmed-up match could not explain that experience, so this pass measured first and repeated casts in fresh Mac game processes.

Before the clock starts

The game now briefly shows “The mire is waking” while preparing the selected player and rival’s visual effects in the actual arena. Spell warnings, active effects, arrivals, hits, splinters, scorch and water responses are covered. The match clock, movement and combat stay frozen until preparation finishes.

Crimson’s heart and Tempest’s coil also remain loaded instead of being read again during ritual creation. The ability inspector uses the same antialiasing as the game. None of this changes spell prices, damage, cooldowns or the eighteen-soul economy.

The short dark loading veil with the words The mire is waking.
Actual exported loading screen. Automated held movement and fire leave the match unchanged during this phase. A separate check deliberately disables automatic window drawing; preparation still completes and normal pause, restart and menu return work.

What the measurements say

The original Crimson rituals commonly spent 35–54 milliseconds just constructing their effects. Retaining their resources reduced that work to a few milliseconds. In a separate Tempest comparison with initially empty Godot shader caches, first Heavenfall construction fell from 46.5 to 1.9 milliseconds. The largest frame in that cast’s observation window fell from 92.8 to 26.8 milliseconds.

Those comparisons do not promise a hitch-free game. An early baseline included a one-second frame that did not reproduce consistently, and occasional long frames also appeared in a no-cast control. Existing driver caches, desktop state and display pacing limit the comparison. The underlying first-use work is addressed; Shaun’s hands-on replay remains the next test of the reported experience.

A preserved checkpoint, then the bodies

Both sides of all four spell families passed first and repeated paid-cast probes. The exported build also passed loading, held-input isolation, menu and saved-control checks, terrain reactions, restart, asset inspection, standalone startup and archive verification. Loading is explicitly rendered so a window that stops drawing automatically cannot leave it waiting forever.

Direct desktop interaction remained unavailable while the Mac reported locked. Native Metal rendering and scripted input supplied the captures and checks; human art, audio and control acceptance remains separate.

The previous complete build is preserved. Next comes the agreed Crimson presentation polish, retaining the generated bodies, followed by Vessara, Drowned Tempest and Vorr actor upgrades within the existing Meshy budget. No additional Meshy credits were spent on this repair.

End of field note 29Back to the journal