I am trying to determine whether or not this is an execution plan reporting bug or an actual feature that is working as intended.
In my execution plan, when using MAXDOP 1, I am seeing the entire table being scanned (https://www.brentozar.com/pastetheplan/?id=HkPFquLdc - Very bottom of plan Object 11 shows ~2.5 million rows read which is the entire table).
However, when I let the engine choose it's own plan with no hints, it goes parallel (https://www.brentozar.com/pastetheplan/?id=r12p5OUdc) and does a bitmap/hash match and that same Object11 only shows ~534k rows read despite doing an index scan and having a bunch of predicates on other columns not in the clustered index.
I would expect that SQL has to read every single row in the table to evaluate each predicate, but perhaps the PROBE IN (you can't see this in paste the plan) on Object 11 in the parallel plan is able to "Filter out pages" since the probe is on the PK/CX.