Tests fail because of top-level ids now beeing the same, xref #25158. Currently those tests pass because of === fallback comparing two identical frames, which seems brittle too. Should the patch in this PR instead generate a id from a hash of the frame or something?

Can you explain briefly what issue this will fix? The description above is blank. Re-running CI now, since we recently suffered a loss of all logs.

I had a log record struct with ::Symbol on the message id field but that broke for deprecation log messages. Perhaps the assumption that the id is always a symbol was wrong though?