the ?passing filter fails during cluster upgrade from 1.18.2 to 1.22.3

#### Overview of the Issue

The breaking change — moving ?passing evaluation from a client-side post-filter on the follower to a server-side field on the request — was shipped silently as an internal refactor, with no upgrade note, no breaking change callout, and no mention of the mixed-version incompatibility it creates. Given that 1.18 is an Enterprise LTS release that many people would be upgrading from, this is a pretty significant omission and a strong case for a bug report against HashiCorp.

---

#### Reproduction Steps

### Consul info for both Client and Server

server=true

```
Output from client 'consul info' command here
```

```
Client agent HCL config
```

Server info

```
Output from server 'consul info' command here
```

```
Server agent HCL config
```

### Operating system and Environment details

### Log Fragments

Start by reproducing the mixed-version cluster described in the issue: upgrade follower servers from Consul 1.18.2 to 1.22.3 while leaving the leader on 1.18.2, then query /v1/service/health/dummy?passing. Compare the returned instances with their health status; done means the filter behaves correctly during the upgrade and the compatibility issue is covered by a regression test.

Written by the indexing model from the issue text.