A longer walk through the mire.
The parish has had a longer run: ordinary play, sixteen-creature combat and a separate twenty-four-agent procession. The Mac checkpoint is checked, archived and ready for the next visual pass.

With the characters, arena and controls in place, this iteration gave the complete scene time to run. The aim was to check frame delivery and repeated effect cleanup, then preserve a dependable Mac build before refining the interface and atmosphere.
Two minutes at a time
Three native Godot workloads each ran for two measured minutes after warming up, rendering a 1920×1080 viewport on this Apple M5 Mac. Ordinary scripted play averaged 117.8 fps; sixteen-creature combat averaged 104.7; the twenty-four-agent stress route averaged 101.0. Their 95th-percentile frame intervals were 9.33, 10.11 and 10.46 milliseconds respectively.
The Mac was locked during these measurements, with rendering continuing into the window. They describe this machine and these fixtures, rather than a minimum hardware specification or physical-display acceptance. GPU timestamps were unavailable. A separate short exported-pack smoke recorded one 43.8 ms frame interval, so this is not a claim that every frame stays above sixty fps.

More feet, fewer leftovers
The stress fixture sends twenty-four paid Puslings around the real navigation routes. It temporarily expands its own cap and soul ledger. Normal matches still allow eight creatures per side and eighteen souls in total. Every stress agent moved throughout the measured route, and all ledger checks passed.
Repeated-effect checks attempted 960 impacts, 96 projectiles and 24 area spells. The effect limit held, expired effects cleared, and the warmed scene finished with stable node, resource and tracked video-memory counts, plus zero orphaned nodes. There was no demonstrated bottleneck requiring a game rewrite.

A preserved Mac checkpoint
The final full regression passed 681 checks, with ten audio-file checks alongside it. The export passed 114 dependency and exclusion checks, 54 workload checks, standalone startup, signature and archive verification. After Shaun unlocked the Mac, computer use checked the actual app, settings, restart and pause.
Version 0.5.5 is archived with the previous build preserved. Its entire game pack is byte-identical to the already tested 0.5.4 pack: the release adds bundle version metadata and development verification, without changing gameplay. Windows, physical Intel execution and subjective sound or combat acceptance remain unverified.
Shaun reports that performance feels fine. The next iteration brings the HUD, health bars and abilities closer to the original Lantern Mire concept, then adds low swamp mist and deeper atmosphere while keeping targets and routes readable.