Clarification needed for `sample-rate` configuration

The docs mention that reducing sample-rate to f.ex. 0.1 would mean that randomly only 10% of errors would get forwarded to the server.

Is it really random for all events or does it, for example, forward each first event and then after that only forwards every 10th event?

If it’s just randomization for each event then it seems to me that we run a risk of missing out on rare case bugs etc.

It’s really random yes.
And yes you are right, in theory, you could miss an important event but it works surprisingly well if you have a reasonable amount of events sending in.
You will still catch the important bugs.

It’s meant to be a simple way of doing sampling on the client-side for high volume scenarios.