Direct answer

Measure a raid while waves are spawning and moving, not only while the farm is idle. Separate raider AI/pathfinding costs from item collection and redstone pulses.

How to check it

  1. Record the number of simultaneously active raids, nearby players and raiders. Use a comparable idle baseline, then run /spark profiler start --timeout 300 during the active raid on supported Paper versions.
  2. Look for time in entity ticking, navigation or block updates. If the farm uses villagers, also inspect point-of-interest and villager activity instead of assuming every spike is caused by raiders.
  3. Run a short A/B test with the raid paused or reduced, while preserving the rest of the farm infrastructure. Compare MSPT and the timing of each raid wave.

What to change

  1. Avoid retaining large groups of live raiders between waves. Improve how quickly the farm processes already-spawned mobs without designing a system that depends on unintended exploits.
  2. If the collector backs up, reduce loose item accumulation and test hopper/clock activity independently; these may be the actual bottleneck.
  3. Do not blindly reduce AI settings: raider movement and raid mechanics can change, breaking the farm or its intended difficulty.

Software and version notes

Version and mechanics warning

Raid behavior and farm designs depend heavily on Minecraft release, difficulty and server implementation. Keep mechanics unchanged while locating the measured cost.

Verify the result

Repeat a complete raid under comparable player load. Report peak MSPT, average MSPT, wave completion and loot output before and after.

References