Adithya Kumar

The regression test was across `9.8` and `9.10` IIRC.

Harendra: Check GHC 9.10 perf regressions. We can use bisection to find which change in GHC caused the regressions.

`maxThreads` defines an upper bound on the number of threads spawned. If you're seeing a lesser number of threads, it may be the case that `compileCssForSingleClient` completes before other threads...

Consider renaming intersperse-ish combinators. ``` intersperse -> sepEachBy, sepPairBy? intersperseEndBy -> endEachBy, eachEndBy? ``` Or consider keeping intersperse as is and rename other flavours to `endEachBy`, `beginEachBy`, etc.

We can close this issue if not required. This is for tracking `Streamly.Internal.Data.Stream.Async.mkAsyncD`

These `TODO` statements are not required immediately. We can do them lazily when/if required.

In the stream function drivers, the abs-position argument is at the end. We should put it as the first argument just because stream and the stream constructors do the same....

Benchmark regressions need to be checked.

Hi @clintonmead , thank you for the detailed comment. I can see one problem with the implementation of `limitParserInput`. It does not account for any backtracking. In the current implementation...

A custom driver might not be required, if we can save the buffer in the parser combinator itself, the parser combinator can do the job of the driver. ``` backtrackSize...