I am working on a tool which will make it possible to search edit summaries. The basic idea is to parse the page.revision.comment tags out of the pages-meta-history SQL dump files and build an elasticsearch index for them.
Details are still unclear, but it's looking like the dump parsing and search index building will both take place as k8s batch jobs on Toolforge. My initial work will be a proof-of-concept using a small subset of the dump data. Once I've got that working, I'll be in a better position to estimate the resource consumption for the full search index.
The original incentive for this is for use in sockpuppet investigations, so I anticipate this will be exposed to end users via the spi-tools web service running on https://spi-tools.toolforge.org/
I run spi-tools as the production service, and spi-tools-dev for testing. It would probably make sense to have distinct elastic search credentials for each of those.