WT-15818 proposes crashing generally when reading ahead of the materialisation frontier. We're not ready for that yet, however in the specific case where we're a leader and the LSN isn't a "special" one, we don't think that path should be reachable.

Even though we think it shouldn't be reachable, it does seem to happen sometimes in prod. This warrants something a bit more involved than just the current warning log line. Some of the ideas floating around are an assertion, upgrading the line to an error message (so we start failing tests when it trips), and printing a backtrace when it occurs.

The scope here is to (1) pick from the available options and/or do something else, and (2) get it implemented (+tested).