All of their public addresses are announced via the same AS#, but because the legacy BGP table is so bloated and messy you can get corner cases where routing is different - especially if the announced prefixes are different sizes. Some of the peers may filter smaller prefixes, or drop excess numbers of prefixes etc - which will then cause the traffic to take a different route.my ping to cloud flare still 30-40ms. With no real solid evidence (just my own ep, It feels like Singtel has different routing for each of their different public IPs for whatever reason. Own (admitedly not scientific or controlled) exp was that I used to have lower ms pings to cloud flare, but ever since I got a new IP its mbeen 30-40ms and routing to certian Local game server (namely Apex Legends SG server) has been severely affected. (talking about 200ms ping or ping spikes randomly up to 200ms)
Also take a look at the prefixes section of https://bgp.he.net/AS9506:
https://bgp.he.net/AS9506
You'll see they have multiple overlapping announcements, where some of the more specific announcements have low visibility. Also a lot of their v6 announcements have invalid RPKI. All of this will cause some peers (but not all) to reject the invalid routes, causing traffic to those destinations to reroute via peers that don't reject it - which may be slower. You may find you have an address which falls into one of these low visibility routes.
Or it could be the other way round where the low visibility routes are intentional to enable peering with specific destinations.
For v6 there is an overall /30 route which covers everything, so peers can aggregate to just this and ignore all the others.