Bugen Zhao

> can we use stack_trace to add span for notify Yes, but this should be treated as the fallback solution. It does become easy to debug deadlocks or stuck in...

> Why not just mark it `#[cfg(test)]`? Yes, it's also okay to mark `#[cfg(test)]`. For some cases, the code might be temporarily unused by production and will be activated after...

> Atomicity can be guaranteed by not injecting barrier in between these chunks. This might be problematic if the query is too large. 😢 I'm considering persisting the input chunks...

BTW, there're some more serious consistency problems unresolved for `update` and `delete`. - https://singularity-data.quip.com/KlRvAysabUma/Handle-Concurrent-Writes-Correctly - https://github.com/singularity-data/risingwave/issues/1186

Really want https://github.com/rust-lang/cargo/pull/11114 after bumping. 🤤

Seems there's some new false positives for resolving lifetime failure. 🤣

Will implementing state table iterator manually with `poll_next` help?

Actually, the interface provided by the storage engine is actually with the "upsert" semantics... and the `update` interface, which does nothing except for the sanity checking, did not exist initially....

> e2e tests do not seem to cover this change, related to https://github.com/risingwavelabs/risingwave/issues/5080 The newly introduced scaling test will cover this. :)

Seems like there're some issues to be resolved. Would you please open a tracking issue for this? Like #1380.