Currently, multiversion tests can fail unexpectedly in case of backports. A test on a 'newer' branch (for instance v7.0) can fail with little indication that the reason of the failure is the new release on an 'older' branch (for instance, 'last-lts' changing from 6.0.19 to 6.0.20).

I believe such failures do not happen very often, but when they do, they could be a bit annoying.

Storing the information which version is the 'last-lts' for the v7.0 branch in the repository would make it a bit easier to trace such version changes (for the price that someone has to 'bump' the version manually).

A related problem is when the tests start to fail because of the changes in the jstestfuzz repository.

Regardless of where the version is stored, the difference between 6.0.19 and 6.0.20 is 65 commits and it can be more than that. In other words, we might notice an error much later after it was introduced. As a possible solution, we can introduce additional 'reverse' multiversion-only regular test runs that compare current v6.0 against v7.0 'last-lts'.