e.g. is AS18566 the origin AS for 751 prefixes that could be collapsed to 6?
Sort of - from here it looks like they aren't actually announcing
The covering /162, /15 and /14 aggregates are being globally announced, and the more specifics are being announced from the same origin AS, apparently along the same AS paths as the specifics, so its not clear that there is any form of traffic engineering or load balancing going on.
o and who can hack the perl to generate filters for this so we can listen only to the aggregates
I don't see their stuff listed in the radb, so it looks like you
have to go to the router itself to read the routes to determine what they are announcing, then ask the appropriate registry which block they are announcing from, aggregate the routes in that block, rinse and repeat, aggregate the aggregated blocks, then filter. ..but that only works if they announce the supernets. Without that I'm not sure what you can do that wouldn't end up biting you back. It's also a lot of work for not too much reward as long as these messes are the exception and not the rule.
Plus the public hand slap is kind of amusing. But that's probably