We can’t reproduce this internally, and we don’t see a corresponding bump in the telemetry we have on raycast performance.
It would help us if anyone could supply a Microprofiler dump that shows a really long raycast, or supplied a place file that consistently shows the problem.