This ticket is an excellent example of a terrible ticket.
1) It does not lead with a problem statement. The title takes the form akin to "I want X". It proposes a tactic.
2) The description is woefully inadequate. Here too, no problem statement. Descriptions should point to any discussions, but discussions are long and rambling, and no substitute for a succinct problem statement in the description. Descriptions need to be maintained, with the current strategy and name of best patch.
3) No resolution strategy. Just patches attached. How is a reviewer supposed to assess your patch if you don't even bother stating what the problem is and what your approach will be in fixing it, how that approach does fix it, and what if any tradeoffs are being made?
4) The change being requested would be a breaking change. That should be made extremely clear in the description, and doubles the threshold for quality of motivation and implementation.
5) Any breaking change would have to carefully enumerate what would break and when, what the migration strategy would be etc.
6) The patch breaks the compiler. If you don't understand how the compiler works, and why features are there, you shouldn't submit patches that alter it. The only assessments made in comments are 'it works for me', which, while useful anecdotes, are insufficient.
While the ticket itself was bad, the uncritical rallying behind it was more troubling. This is not how Clojure was built, and will not be how it is maintained.
In the end, the ticket proposed a tactic, and that tactic has been rejected. Read safety has been addressed by other means.