I've never heard dry-run taking that much time but I believe it can happen with complex enough data. Not validating keywords that have already been validated sounds like a good idea, but I'm not sure how easy it would be. Some challenges listed below:

• How to reliably that detect a keyword in one suite is same as a keyword in another suite? Keywords and even resource files where they are can have same names. Resource files having a references to the original source path would help but I don't remember do they have it. Could definitely be added if not.

• What should actually be done if we detect that a keyword has already been validated? Is it enough to only add a note to the log about it or should the whole validation be copied there?

Problems above are likely solvable and prototyping this would be great. I personally won't have time for it in the foreseeable future, but perhaps you @myjniak could look at it yourself. Or perhaps @mkorpela would be interested because, AFAIK, pabot also uses dry-run internally and would benefit from speedups.

Pabot internally hacks away most of the data in dryrun as pabot is only interested about the names of the test cases that will be executed.
This was actually done because some people complaint about the operation taking so much time. And the hack improved the situation.

No updates since 2019 so this doesn't seem to be too critical. New issue can be opened, or this one reopened, later if needed. We'd most importantly needed a PR implementing this.