Lyn Nagara

I personally far prefer option 3 over 1. I think 1 can end up hurting productivity across all of our repositories that are shipped as part of self hosted (which...

What is the motivation for this? Does it tell us anything that the metric doesn't already?

can you provide some context about the use-case for this? afaik it's generally used for batching which isn't a backpressure scenario. i believe the python version of this strategy also...

Were you able to confirm all sentry and snuba consumers have adequate memory requested for this?

Can we sample aggressively? Approximate trends in sizes may still be useful? Though not sure why it's slow in the first place.

Have you seen https://github.com/getsentry/snuba/blob/ce8470829bc683345695bcd3ef26492757727c14/tests/consumers/test_schemas.py? It attempts to run all message processors against the example payloads in sentry-kafka-schemas, including spans. Not sure if this overlaps with the intention here, or this...

> Does this require that we put stuff in the examples folder? the only examples we have so far are the ones in this test, but I can craft some...

there's a lot of places in code where you need to implement the dual writing, have you considered putting the dual writing logic in one place only inside the `transaction_processing_store`...

i think there are a few more places in the codebase where this is needed.. e.g. https://github.com/getsentry/sentry/blob/939be86fe916546f6020dcddeba2f5207071e8d8/src/sentry/event_manager.py#L1104-L1107 can you check the whole codebase?

> For my education, why is this safe? It'd seem that this restricts the endpoint to control and thus old UI code would be potentially broken until a new UI...