Hans Svensson

This is the category of changes that are more or less trivial to do implementation-wise, but where reaching a consensus is the hard work. Sounds like something that should be...

@marc0olo Nothing can be done short term since the auction is part of the consensus protocol. So ~~hard-fork~~ protocol upgrade needed. Different schemes with regards to auction extension and payback...

I think I see what goes wrong. But not why and how to fix it… What happens when we build a prod-package is that our beloved `rebar3` (and internally `relx`)...

We are not building with an old `rebar3` - the problem turns out to be that step where `relx` copies `dyn_erl` over `erl` is allowed to silently fail (and it...

@tsloughter confirmed; the copied `erl` is owned by the current user but read-only

@tsloughter Any luck fixing this, I had a look over at `relx` but didn't see anything related?

Isn't there exactly the same issue with pointers (and client_ttl) - where we also sometimes pass `None` and keep the previous value?

For`client_ttl` I guess `-1` would work. But for pointers I don't know, there are no type correct values that are easily distinguishable 🤔

412 is some caching stupidity, let me guess you are not using a simple node but something behind a loadbalance/cache?! I see Cloudfront being mentioned in the log - I...

Note: This was not my idea, I'm merely noting it down here to keep a consistent record of possible CERES additions. It sound seemingly simple, but I wonder if the...