Wallets are incentivised to do a better job with defragmentation already, as if you have lots of tiny UTXOs then your fees end up being huge when trying to make a payment.

The reason they largely don't is just one of manpower. Nobody is working on it.

As a wallet developer myself, one way I'd like to see this issue be fixed by making free transactions more reliable. Then wallets can submit free transactions to the network to consolidate UTXOs together, e.g. at night when the user is sleeping. They would then fit into whatever space is available in the block during periods of low demand, like on Sunday.

If we don't do this then wallets won't automatically defragment, as we'd be unable to explain to the user why their money is slowly leaking out of their wallet without them doing anything. Trying to explain the existing transaction fees is hard enough already ("I thought bitcoin doesn't have banks" etc).

There is another way: as the fee is based on a rounded 1kb calculation, if you go into the next fee band adding some more outputs and making a bigger change output becomes "free" for another output or two. But wallets don't exploit this today.