When feerates are high we may fail to spend outputs (eg low-value htlcs or to-self outputs), which currently unwrap's. We really need a pipeline to store them to disk and claim on a timer anyway in case fees go up after a claim, but we at least shouldn't unwrap.