Unless they’re properly compensating you, and I mean something like a revenue share on that mess if you fix it, I’d walk if I were you. The job is high risk, is

Unless they’re properly compensating you, and I mean something like a revenue share on that mess if you fix it, I’d walk if I were you. The job is high risk, isn’t properly staffed, doesn’t have management support, and sounds intrinsically miserable. They’ll balk I’m sure, but point out to them that their only real alternative is contracting high end consultants who will demand even more. If you do manage to get them to give you some points of the revenue from the project, then start with introducing source control and follow by building up a testing process and integration environment. I’d probably use tcpdump to capture a week’s worth of live traffic and replay it with a suitable speed up to replicate testing in production in your integration environment. That should give you serviceable integration tests. To start by writing unit tests will be pointless because it sounds like there are no discrete units to test. From there you’ll want to apply some kind of strangler pattern where you incrementally replace pieces of the system. Doing that will require some refactoring to start separating concerns. Again don’t try to do it all at once and don’t try to make it perfect. Then you can start introducing unit tests. Then there’s the database, which is a full size job in its own right. And who knows what other unpleasant surprises await, but bank on them being there.