Yes, as I mentioned in my other replies, what I've seen is that quality has gone up and coalesced around a much higher bar with far less variance than before as we've refined our docs and tooling.

In some cases, it was "instant"; dev's MCP server connected to our docs was down -> terrible PR. We fix the MCP connection and redo the PR -> instantly follows the guides we have in place for best practices.