This costs.
In any case a couple of wallets is enough.
The problem for the attacker is that they have to wait for the coin to be sent. We say payments are instantaneous but obviously the reality is that it will take a finite time, maybe a few seconds to clear the transaction is more realistic.
If the attacker tries to send the coin on before it has cleared (ie trying to spam the coin) then there is a chance for that coin to be lost as David points out (somewhere). The potential for losing the coin only happens when one tries to spam or attack the network by “double spending quickly”
Now if the attacker has 1 million coins being spent at 3 seconds put transfer, then maximum transaction rate by the attacker is 333,333 transactions per second across a world wide network. Now for a small network (early on) this would result in transfers taking even longer reducing the effects of the attack.
Testing is obviously needed but the delays in doing the transaction make this kind of attack less effective. The same is true for division if my method is adopted as it uses the same basic mechanism for sending.
The ultra important point here is that unlike a blockchain any attack like this doesn’t clog up the whole system, each transaction is its own entity and isn’t being confirmed by energy/processing heavy processes across the whole network, but simply consensus by two or 3 groups.