When we start up, we currently fetch all state across all rooms that we are joined to. This is ludicrously expensive, but reasonably safe to do. On large deployments this effectively means there is a noticeable gap between restarts. We can do some tinkering around the edges here by caching the rooms which are "active", and do fewer state requests as a result.
When we start up, we currently fetch all state across all rooms that we are joined to. This is ludicrously expensive, but reasonably safe to do. On large deployments this effectively means there is a noticeable gap between restarts. We can do some tinkering around the edges here by caching the rooms which are "active", and do fewer state requests as a result.