Connection worker delay is not being respected

[steve-chavez]

On v7, it used to be that all requests would get rejected with 503 while the connection worker is running. But on v8.0, the connection recovery is done whenever a new request comes in:

curl -I localhost:3000/projects
HTTP/1.1 503 Service Unavailable
Date: Tue, 17 Aug 2021 17:15:12 GMT
Server: postgrest/8.0.0 (b3899e7)
Retry-After: 32
Content-Type: application/json; charset=utf-8

sudo systemctl start postgresql

## just a second after

curl -I localhost:3000/projects
HTTP/1.1 200 OK
Date: Tue, 17 Aug 2021 17:15:46 GMT
Server: postgrest/8.0.0 (b3899e7)
Content-Range: 0-58/*
Content-Location: /projects
Content-Type: application/json; charset=utf-8

## postgrest is already recovered

## The connection worker will report the reconnection later(after 32 sec):
## 17/Aug/2021:12:16:05 -0500: Connection successful
## 17/Aug/2021:12:16:05 -0500: Config re-loaded
## 17/Aug/2021:12:16:06 -0500: Schema cache loade

This is probably a side effect of doing #1559.

[wolfgangwalther]

Not sure what the dev label is supposed to mean here. I might not fully understand the issue, but this looks like a regression from v7 to me - so a bug?

[steve-chavez]

Edit: Left for later

Will work this on #2813. But even with no change I believe it's not a bug(it's actually a feature) since the Retry header is just a hint - we can recover earlier and this is better for all clients.

---

Not sure what the dev label is supposed to mean here.

Sidenote: I also don't get that label(I don't recall adding it). Will delete it and move the usages to "hygiene".