Yep, I saw concurrency and delay setting for the active scan, but it's not exactly what I'm looking for. For example, when scanning APIs, sometimes you have a quota for requests sent per minute.
If a quota is, let's say, 60 requests per minute and ZAP exceeds this limit, the rest of requests will be simply blocked by an API. Since this quota may vary among different APIs it would make sense to have this setting per host.