Probably spun up a new node with an empty mempool.dat. Or alternatively, deleted their mempool.dat on an existing node.
Currently Bitcoin Core lacks a mechanism to sync your mempool with your peers at startup. Given the amount of money involved, this might be worth implementing.
Automatic rebroadcast from the mempool could help somewhat, though not as well as being able to pull from peers in this situation. The idea in the PR below is both reactive and initiated by the network rather than the miner. It was closed and I can't tell if it's being worked on elsewhere. It seemed like a good idea and another piece of the puzzle, I was looking forward to it.
Are there privacy risks associated with a peer serving its mempool to others?
Yes. There are ways to mitigate those risks too. The main privacy risk is related to learning about propagation, so providing only old txs that multiple peers have tried to give you would probably be a good mitigation. You could also have nodes opt into mempool downloads, so only a subset would provide this service. Plenty of other mitigations too.
Main thing is now that having a missing mempool is so expensive, it's arguably work putting in the engineering effort to improve this. Previously that was a harder argument to make.
Yeah, there are privacy concerns and bandwidth usage concerns. As far as I know work on this particular approach has been discontinued.
The latest related idea I heard about is to do initial broadcasts per short-lived connections to nodes on the Tor or I2P networks. If adopted, this approach could perhaps also be used to perform rebroadcasts without leaking the sender’s IP address.
savemempoolandimportmempool.