kurin
Update the docs to reflect the use of context for handling overruns.
A bunch of Backblaze pods had their SSL certs expire, and b2 kept retrying until the user got annoyed and quit.
Should this have been a permanent error?
I think a TLS cert error could sensibly be considered fatal, as with any other error that suggests that the remote end is active but misconfigured and thus unable to be able to serve the request in the near future anyway.
For other things, the spin forever should probably just be dealt with through a user providing a context with timeout.
I think the context in which this issue came up was that someone had issued an interactive command in a third party binary (like list backups) and then it just hanged. In a scenario where you're actually writing to b2, I would want blazer to spin and retry.
I think the right way to handle this is for the caller to set context timeouts appropriately for interactive uses.