Direct answer
Folia must budget chunk-generation workers separately from region tick threads. Pre-generated worlds usually need less chunk-generation capacity than exploration-heavy worlds, and adding workers reduces CPU headroom for region ticking.
How to check it
- Measure exploration/new-chunk generation separately from idle and pre-generated-world performance.
- Note physical cores, GC workers, Netty I/O, chunk system I/O and configured tick-region threads.
- Inspect the running Folia release’s global configuration for the actual chunk worker and I/O keys; do not transplant Paper YAML blindly.
What to change
Folia’s FAQ suggests rough historical starting allocations for pre-generated environments but states that exploration-heavy cases have no reliable one-size-fits-all worker count. Prefer pre-generating the world if that suits your server, then tune workers and region threads separately.
See Folia Tick Threads for a broader CPU-budget explanation. Avoid attempting to allocate every core simultaneously to region ticks and chunk workers.
Software and version notes
The Folia FAQ explicitly calls its ~330-player test estimates imprecise. Do not treat those example thread counts as current defaults or apply them to unrelated server sizes.
Verify the result
Test pre-generated travel and newly generated exploration separately. Confirm that chunk wait time improves without raising region MSPT or overall CPU contention.