For "EU Portal API" this is just a reference link shown on the Sources page — the actual fetching uses the portal's own public API, not this URL.
The portal currently lists roughly 750 open/forthcoming calls. Results are fetched soonest-deadline-first, so a lower cap trades off the calls with the most distant deadlines first — the near-term ones are always fetched before them.
Each candidate costs at least one AI extraction call, which is what makes a run slow — lower this for a faster run, or raise it (up to "Results to fetch per discovery pass" above, for eu_portal sources) to clear everything not-yet-seen in one go. A run saves its progress as it goes, so a large batch is safe to leave running or interrupt — nothing above the cap, or lost to an interruption, is gone for good: it's picked up on the next "Check now" (already-known calls are always skipped).