[SalesForce] Apex class ‘SiteLoginController’ does not exist (scratch org)

I've created over 40 scratch orgs with the following config, without issues.

{
    "orgName": "MyScratchOrg",
    "edition": "Developer",
    "features": ["Communities","PersonAccounts","SiteDotCom","ContactsToMultipleAccounts"],
    "orgPreferences" : {
        "enabled": ["S1DesktopEnabled","NetworksEnabled"]
    }
}

Today I tried to spin up a new one (actually two, one after the other), everything went well, but the deployment failed when deploying the community/network/site stuff with the following message:

Apex class 'SiteLoginController' does not exist

This is a class that should be automatically created once you enable communities right?

So far I've only found this

Salesforce CLI Sometimes Doesn’t Recognize Scratch Orgs with
Communities Description: Sometimes, but not in all cases, the
Salesforce CLI doesn’t acknowledge the creation of scratch orgs with
the Communities feature. You can’t open the scratch org using the CLI,
even though the scratch org is listed in Dev Hub. Workaround: You can
try this workaround, although it doesn’t fix the issue in all cases.
Delete the scratch org in Dev Hub, then create a new scratch org using
the CLI. Deleting and recreating scratch orgs counts against your
daily scratch org limits.

Error Occurs If You Pull a Community and Deploy It Description: The
error occurs because the scratch org doesn’t have the required guest
license. Workaround: In your scratch org definition file, if you
specify the Communities feature, also specify the Sites feature.

In my case, I've got enough licenses, I've specified both Communities and Sites, and I can actually login to the scratch org, therefore this limitation is probably not the culprit…

Anyone else experienced this? What else could be the issue?

Update:
I was able to create a third scratch org today, this time without issues. I'd still like to get to the bottom of this and find out what causes this intermittent issue…

Another Update
I tried to run 2 bitbucket pipelines in parallel which create scratch orgs (to test the theory of possible issue with parallel creation/deployment), however this time around one succeeded, whereas the second threw yet another new error:

'ENOTFOUND': getaddrinfo ENOTFOUND

Which I guess might be related to the subdomain not yet propagated through the DNS servers… In any case, I'll be opening a case with support…