Comment by kagen

That’s a good question. The plan depends on the enclosure version: whenever changes are made to an enclosure, its version is updated, prompting every client to recalculate the plans for all animals. To avoid abrupt jumps, an animal enters a special, short-term "return-to-path" mode—moving from its current position to the location dictated by the new plan—before resuming its standard movement. Since different clients receive the update via SSE at slightly different times, their data may diverge for a second or two before synchronizing back to a consistent state.