Evan Huus

This explanation makes sense to me, though it raises one other question: I _consistently_ see `execute_multiplex` running _after_ lex and parse, and you seem to consistently see it running before....

This is in core, which instantiates a `GraphQL::Query` instance with the string, and then calls `.result` on it. AFAICT this intentionally lexes/parses before it enters the multiplexer: https://github.com/rmosolgo/graphql-ruby/blob/5c3a8ec6f45319f40cb59f2ffb5961bb1071aa80/lib/graphql/query.rb#L198-L200 This may...

The motivating use case makes sense to me. The relevant line in the spec I see is > If an error is thrown while resolving a field, it should be...

Yes, you are correct.

Hmm, to be honest it's the constructor (`New`) which takes the static list of backoff values. If you wanted to design a proper dynamic retry interface, it would involve a...

Yes, *if somebody is interested in working on it*, then that larger discussion should be happening in a different issue or PR at this point - the ideas being floated...

I'm going to leave this PR open for the time being - this is a reasonably small extension of the existing implementation that doesn't really get in the way of...

I think something like that interface could make sense, yes. As discussed though, switching from `[]time.Duration` to something like that starts to impact the constructor and overall design of the...

This rings a vague bell for me, but I don’t remember concretely, and I’m unfortunately no longer at Shopify to check. I believe @swalkinshaw or @rebeccajfriedman can probably confirm this...

I'm a bit confused by "Input Value Uncoercion". If it's deprecated why is it part of the ideal end state? Also I'm not sure how it's meaningfully different from "Result...