Are in-place Windows Server upgrades worth doing now?

For ordinary application and infrastructure servers, in-place upgrades are a reasonable option now. They can save a lot of time reinstalling software and reproducing configuration. That doesn't make them a non-issue, though. Check the supported upgrade path and application compatibility first, especially for older IIS features or other dependencies that may have changed.

Have a tested backup and recovery plan, schedule downtime, and verify the application and its dependencies afterward. A VM snapshot can be a useful short-term rollback option, but I wouldn't treat it as a replacement for a proper backup.

I'd make domain controllers the exception. Microsoft recommends a clean install and promotion of new domain controllers rather than an in-place upgrade for AD DS when moving to Windows Server 2025 or later. An in-place upgrade is technically possible, but Microsoft says it doesn't deliver the AD performance and feature improvements included in those releases.

Before retiring an old controller, check for devices with hard-coded DNS or LDAP addresses. Those dependencies can make an otherwise straightforward replacement more awkward.

I still lean toward a fresh build for servers with a long history or poorly documented configuration. You get a chance to clean that up, and an upgraded OS is one less variable to investigate if an application has trouble later. It's more work upfront, but that work isn't necessarily wasted.

Don't make it an all-or-nothing policy. Pick a suitable low-risk server, test the upgrade and recovery process, then decide what to do with the rest. If a server needs a change from BIOS to UEFI, or is a cluster node that can be replaced without downtime, a new build may make more sense.