[1.20.1] Severe hitch caused primarily by world-generation logic when entering a world #175

Summary

There is a severe multi-second hitch when entering a world. The primary issue appears to be the world-generation pipeline, especially the Starter Structure generation logic. EMI recipe/item-group loading is a secondary contributor.

Environment

• Minecraft 1.20.1
• Forge 47.4.10
• Java 17, aarch64
• macOS Apple Silicon
• Compression

Reproduction

1. Launch the pack.
2. Create or enter a world.
3. Observe the long pause during the transition into the world.

Primary evidence: world-generation blocking

In an earlier test session:

• Preparing the overworld start region began at 04:11:22.
• Progress remained around 9% for a long time and only completed around 04:12:05.
• At 04:12:17, Starter Structure began generating a structure containing 10,129,231 blocks.
• The integrated server then fell behind by 14,821 ms / 296 ticks.

This indicates that the main stall is caused by world-generation logic doing an extremely large synchronous structure operation near world entry.

Secondary evidence: EMI and recipe loading

The latest test session also shows:

• EMI reloaded item groups on the client thread for 1638 ms and then 734 ms.
• Recipe generation failed for entries using create:sandpaper_polishing.
• Other recipe and tag reload errors were reported during the same phase.
• The server later reported Running 2025ms or 40 ticks behind at 05:20:51.

These EMI/recipe-loading problems can add client-side stutter, but they are secondary to the Starter Structure/world-generation stall.

Expected behavior

Entering a world should not synchronously generate millions of structure blocks or block the integrated server for hundreds of ticks.

Notes

The issue should first be investigated in the Starter Structure/world-generation code path. EMI recipe loading and the related invalid recipe data should then be handled separately.