Anomaly in reported crashed sessions

[jjbayer]

A user reported a negative crash-free rate. We figured out that there's a single 10-second interval in which a very high number of crashed sessions was reported, but the number of total sessions remained stable:

This looks suspicious for two reasons:

1. The number itself seems unreasonably high.
2. sessions.all must always be larger than sessions.crashed. If the crashed number is higher, it means that the SDK reported session updates with init: false for which it has never reported the initial update with init: true.

Could there be a reporting bug in the electron SDK which causes large numbers of session updates to be flushed without init: true?

Zendesk ticket with URL to the API request shown above: https://sentry.zendesk.com/agent/tickets/124592

[timfish]

When this issue closes, 5.4.0 will be published to npm. This has a fix for multiple sessions being recorded when there are multiple renderer crashes in a single session. Let us know if this fixes this issue!