Understanding pesticide impacts on populations of target/non-target species and communities is a challenge to applied ecology. When predators that otherwise regulate pest densities ingest prey contaminated with pesticides, this can suppress predator populations by secondary poisoning. It is, however, unknown how species relationships and protocols of treatments (e.g. anticoagulant rodenticide [AR]) interact to affect pest regulation.To tackle this issue, we modelled a heuristic non-spatialized system including montane water voles, specialist vole predators (stoats, weasels) and a generalist predator (red fox) which consumes voles, mustelids and other prey. By carrying out a broad-range sensitivity analysis on poorly known toxicological parameters, we explored the impact of five farmer functional responses (defined by both AR quantity and threshold vole density above which AR spreading is prohibited) on predator-prey interactions, AR transfer across the trophic chain and population effects.Spreading AR to maintain low vole densities suppressed mustelid and fox populations, leading to vole population dynamics being entirely regulated by AR use. Such vole-suppression treatment regimes inhibited predation ecosystem services and promoted pesticide dependence.Keeping vole density below acceptable bounds by spreading AR while maintaining sufficient voles as prey resources led to less AR being applied and extended periods without AR in the environment, benefiting predators while avoiding episodes with high vole density. This may meet farm production interests while minimizing the impact on mustelid and fox populations and associated ecosystem processes. These alternating phases of mustelids and farmer regulation highlight the consequence of intraguild relationship where mustelids may rescue foxes from poisoning. Both global and wide-range sensitivity analysis illustrate the tightrope between predator-prey regulation and pesticide-pest regulation.Synthesis and applications. Different pesticide protocols lead to a rich variety of predator-prey dynamics in agro-ecosystems. Our model reveals the need to maintain refuges with sufficient non-poisoned voles for sustaining specialist mustelids, to conserve the predator community given the potential of secondary poisoning with rodenticides. We suggest that long periods without pesticide treatment are essential to maintain predator populations, and that practices of pesticides use that attempt to permanently suppress a pest over a large scale are counterproductive.