Here are the properties of this idea as I see them.
We end up with a centralized service where:
- if it gets totally compromised => you end up with the same privacy as you have right now
- if it gets ddosed/shutdown => you end up with the same privacy as you have right now
- if it’s working honestly (1/N honest services) => it increases the anonymity set of the outputs significantly (to the point where I’d say we have better practical anonymity than Monero) assuming ~20 txs per hour
The downsides I see are:
- in case of a silent takeover, you’d think you have better privacy than you actually did
- you pay ~2x fees (which is cheap right now and IMO worth increasing everyone’s privacy). If it turns out to be too much, people can opt-out
The great thing is that the service itself has no impact on the actual transactions between the parties so regardless if it works or not, transacting isn’t impacted in any way as far as I can tell. It doesn’t even touch transactions and it only helps with privacy so the only thing that could get hurt is privacy which in the worst case stays as is now. Note that since the transactions are detached from the 1-1 coinswaps in this scheme, they could both be building confirmations in parallel. If the wallets defaulted to sending to the hourly aggregator, the average aggregation would take ~30 minutes and 1 hour in the worst case. In the worst case, we’d wait for our privacy 1 hour which is what Bitcoiners wait for for the confirmations.
P.S. Having a completely trustless service would be very nice. I am however not sure if it’s even theoretically possible to have one that is based on non-interactive aggregation and can handle chain reorgs - which imo kills all hopes of having it by default due to the highly increased damage 
But I guess this is a question for professional researchers and hopefully someone proves me wrong.