Bill Phillips

Can you post this pullrequest to [bignerdranch/recyclerview-criminalintent](https://github.com/bignerdranch/recyclerview-multiselect)?

It's not just Turbine-type interactions that have need to "pull" consumption. We also need it for interaction with fakes. Internally at Cash, we've been attacking this problem by using a...

We have only been concerned with races in test. Test has firm expectations of consistency: many immaterial changes in order may occur in multithreaded prod without harm, but immaterial changes...

> https://github.com/Kotlin/kotlinx.coroutines/pull/3278#issuecomment-1124141560 is an example of code that nondeterministically throws or succeeds, even run on a single-threaded dispatcher. I've stewed on this claim for a few weeks, and reviewing the...

Those arguments make sense to me. But even if I grant all of them, is the conclusion (dispatcher races are dangerous) actionable? Even with the proposed `iterate`? E.g. consider an...

> I think this is the same issue as #2485. Yes, it is. I disagree with the outcome of that discussion, however. > `Dispatchers.Unconfined` would probably not be enough for...

I see. I still disagree, however. Let's call the proposed dispatcher `CallingDispatcher`. The tl;dr of my disagreement is this: _`CallingDispatcher` is better encapsulation, but it does not behave in a...

Assuming my comment was correct at the time I wrote it, then `rxObservable` has been corrected to be consistent with the other tools. That means that I will need to...

Modified to remove the public API to set the `CoroutineDispatcher`. This keeps Kotlin out of the ABI. Instead, always use `Dispatchers.Default`.

Recent updates provide both good-ish and (probably?) bad news on this issue: 1. With the introduction of `RecompositionClock` as a param, there's now support within Molecule for the immediate clock...