Thanks for writing such a useful tool!

In our environment, we use vault to create credentials for stackdriver_exporter. Under certain error conditions those credentials expire and are replaced. stackdriver_exporter then needs to be restarted to load the replacement.

We've currently maintain a one-line fork that accomplishes this by treating a failure to collect metrics a fatal error. This causes our process-manager to restart the service at which point it grabs the new credentials.

We'd love to get something that accomplished the same thing upstreamed. What would you think about a merge request containing this (or a similar) change?