A system may cease performing its present function when one of its modern components is removed, yet that does not prove its ancestors could not have performed different functions with fewer parts.
Evolution does not have to construct the present system by adding components one at a time while preserving exactly the same function at every stage.
This is one of the most important responses to the argument from irreducible complexity. Present-day dependence among components does not by itself reveal the historical sequence by which those components arose.
The objection shows why knockout experiments or removal of parts cannot alone establish that a system could not evolve. Gene duplication, co-option, changes of function, scaffolding, redundancy, and loss of previously useful components can produce historical pathways that are not obvious from the completed system.
At the same time, proposing such mechanisms does not automatically demonstrate the actual history of every molecular machine. Particular systems still require evidence for their proposed pathways.
For systems claimed to be irreducibly complex, what evidence identifies actual ancestral functions and intermediate stages rather than merely possible ones?