The counter to that is what if one agency detects an attack that no other noticed because the culture and practices of that agency diverge from the others in a way that turns out to be positive.

There are both benefits and costs to centralization and standardization. We don't want too much taxpayer-funded redundancy, but some redundancy can be a good thing.

> if different agencies had shared data

Do you think they're not sharing data? Why?

Different agencies is about different collection techniques and domains.

>> the worst attack on US soil.

I think it is pretty strong to suggest that 9/11 was the worst attack on US soil ever without further justification. During the Battle of Gettysburg, for example, more than 7,000 soldiers died after Lee's army attacked the US army. The Attack on Pearl Harbor might also be in contention, depending on how severely you rate the crippling of the US Pacific Fleet.

Does a centralised agency guarantee that this data is shared correctly?

This is different from the Intelligence Apparatus that shares inter-service intelligence. The Intelligence Community is already a thing and is organised under the Director of National Intelligence (https://en.wikipedia.org/wiki/United_States_Intelligence_Com...). The point of each service's Agency is to be good at collecting pertinent intelligence data, analysing it, then sharing it with each other if relevant. This way you're not duplicating Intelligence efforts and getting your best value. Inter-Agency communication is a completely different ball game and should be handled separately from organisational structure.

Maybe, but it always seems like people doing various bad things are known to the agencies and they don't have anything specific enough to act on. So we give them even more funding and manpower to prevent it next time and the cycle repeats.