Investigate "Internal SDK Error" client reports for Replay

[billyvg]

There's a general problem right now where we have replay segments being dropped due to "internal SDK error". We currently have no visibility into these errors, so not entirely sure what is happening. In Sentry, we have the experimental flag to capture SDK exceptions, but it does not seem to be working (it might be related to the DEBUG_BUILD flag).

We need to fix/ensure that replay exceptions that occur on SaaS get captured to Sentry, as well as look into how we can generically capture exceptions from non-internal SDK usage as well.

[bruno-garcia]

@billyvg following some recent investigations on this, do you feel we still need this ticket open?

Is the goal here investigate the Internal SDK Error or specifically capturing Replay SDK exceptions as events in Sentry? I imagine these could be two separate things

[billyvg]

We've added some more logging around this. _experiments.traceInternals option now creates normal breadcrumbs as well so that we have logs inside of our normal events in the case that replays do not get sent/accepted.

The conclusions we've made are that generally, Internal SDK Error means the replay was unable to be sent (e.g. network error, fetch failing, etc).

There's also another case where users are finding that replays are "missing" because they are tracking their users through another service and their sessions do not appear in Sentry. This can be attributed to the minimum replay duration of 5s, which does not get sent.