You are right — because they built their trace data using the same data model as the error data (everything is an "event"), they are now running into difficulties with high quality trace data.

See here: https://develop.sentry.dev/sdk/research/performance/