Write everything (generally, new features) twice has turned out to be really good strategy for me, but it doesn't sit well with bizdev or project managers and tends to be perceived as unnecessary slowness.

Silo-isation compounds this. If the maintenance costs are borne by another team or if any rework will be funded out of a different project, the managers are not going to care about quality beyond the basic "signed off by uat".