Spun pages built from one template with only swapped variables rarely index well, because a search engine sees near-duplicate content carrying no unique value and responds by selecting one of the set or none at all. The mechanism is straightforward to observe: when many pages say substantially the same thing, the system has no reason to store and serve all of them, so it keeps a representative and drops the rest. Scaling pages this way usually scales the duplicate pile, not the indexed footprint.
The reason it fails is what indexing is for. A search index exists to hold pages that offer something worth retrieving for some query. A page that is identical to forty siblings except for a token offers nothing the others do not, so storing all forty is wasted space and serving them is a worse result. When the system detects this near-duplication, it consolidates: it picks the strongest version, or treats the cluster as one, and the rest sit unindexed no matter how many you publish.
This is why spinning fails specifically at scale. One templated page might slip in if nothing similar competes. But the whole point of spinning is volume, and volume is exactly what triggers the near-duplicate consolidation. The more pages you spin from the same content, the more obviously they cluster, and the more aggressively the system collapses them. The tactic defeats itself: the scale that was supposed to be the win becomes the signal that gets the pages discarded.
For your next batch, do not measure success by how many pages you can generate. Measure it by how many carry genuinely unique substantive content that a reader would find useful and no sibling page already provides. If each page passes that bar, scale them. If they would only differ by swapped tokens, accept that spinning them will not index well and either invest real content per page or do not create them at all.