Skip to content


Released: June 20, 2022

Improved Replay Performance

Previously, when calling distributed_materialize_many on high-volume timelines with expensive long-windowed aggregates, some jobs would time out and fail.

This release includes two performance improvements to reduce the incidence of these timeouts:

  1. The query planner now isolates more aggregate features in their own dedicated jobs
  2. Resolution of the feature-dependency graph occurs less often at run-time

As a result, jobs should run faster, in general, and failures should occur less frequently.