Welp. We used it in prod for ~18 months or so but the experience was not something we'd repeat.

The configuration of Traefik, in our case, embedded in the docker-compose file was not clear. What was supposed to be a 'auto-detection' of services ended up looking like a hodge-podge of configs between several files.

The logging was sub-par - we couldn't properly debug issues.

And then we ended up migrating terminating HTTPS on AWS's ELB so the let's encrypt integration became not relevant which catalyzed us going back to nginx.