The Filecoin Proof of Spacetime mechanism is essentially a sampling protocol.
For WinningPoST, SPs are chosen at random using Drand to be part of the set of elected SPs to propose new blocks and thus win block rewards. When a Storage Provider stores LDN FIL+ data, they have 10x the QAP and therefore the chance of getting chosen and of winning block rewards compared.
For WindowPoST, SPs ensure that they are not slashed by providing PoSTs for randomly selected partitions of sectors.
Similarly, Spark is a sampling protocol. Spark checks that data can be retrieved from Storage Providers who store LDN FIL+ data which is assumed to be available for retrieval.
Spark is a trustless retrieval sampling protocol for Filecoin Storage Providers. Spark clients (which can run on Station) form committees that are assigned a SP and CID each round. The Spark clients then make a retrieval request to the SP for the given CID and report their results.
The committee comes to an honest majority consensus on the results of the retrieval.
Similar to a WindowPoST, these results could be submitted on chain as a Proof of retrieval.
If an SP has successful retrieval results on chain, this could increase their QAP to say 15x.