I do think it is a potential problem in theory, and I still think it's good for the error ecosystem to support it.
Honestly though, even though I introduced backtraces into errors in error-chain, I have never used error backtraces for any purpose. I have to imagine that people supporting large codebases would find it useful, as long as they are threaded all the way through the chain, which they almost assuredly will not be considering the fractured state of the ecosystem.