Well then :)

While there is nothing wrong with that patch per se, it would be better to get the sigcheck nailed before landing this.

Good catch, this has been bugging me too.

This doesn't seems to me like this is building capital toward more automation, but this is something.

One small change, but I think this is pretty much good to go.

Would this have detected the perf regression in PR17218? It really doesn't look obvious looking at this benchmark - in fact, I'd bet it's probably not obvious at all. But there are no numbers anywhere, we can all keep guessing.

braces

rebase

The approach is needlessly branchy. You can simplify this.

After more digging, it looks like this is what's core is doing. Let's go for it.

It's all good. With the change we did around NULLFAIL and multisig, we should move things toward verifying signature lazily anyways.

I don't think it makes sense to add a new folder for every tool out there. This dilutes the signal that something at the root is important.

1/ Name it web; www or something.
2/ This needs to reworked. I don't think we want blogs posts in there for instance.

This disallow deleting the native folder and have it be rebuilt.

Let just use an int in the cache and not a ScriptExecutionMetric in the cache.

This introduce several bugs, cf the discussion on the PR.

If the index is a child of tip, shouldn't we invalidate its siblings?

I think the change to accept to memory pool would be better with a static_assert but overall, LGTM.