Background Few real‐world series on the efficacy and safety of anti‐programmed cell death protein‐1(PD‐1)/programmed death ligand‐1(PD‐L1)–based therapy are available in molecularly unselected patients with poor performance status (PS) and specific types of advanced cancers, because such populations are typically excluded from clinical trials due to poor life expectancy and risk of toxicity. Materials and Methods This multicenter retrospective case series included patients with microsatellite instability (MSI)‐high metastatic cancers with Eastern Cooperative Oncology Group (ECOG) PS of 2 or 3 not related to comorbidities receiving anti‐PD‐1 with or without anti‐CTLA‐4 therapy after failure of at least one prior treatment line. Results We included 27 patients with six diverse tumor types: colorectal (n = 18), gastric (n = 5), biliary tract, pancreatic, small bowel, and endometrial cancers (n = 1 each). Baseline ECOG PS was 2 (74%) or 3 (26%). Overall response rate was 33%, with six partial and three complete responses. Median time to response was 3.1, months and median duration of response was 16.9 months. Median progression‐free survival was 3.4 months (95% CI: 2.3 to not evaluable), and 18‐month overall survival was 50.8% (95% confidence interval, 32.7–78.8). Baseline variables were not associated with survival outcomes. ECOG PS 1 was reached by 52% of patients in a median time of 6 weeks, and ECOG PS 0 was reached by 30% of patients in a median time of 10 weeks. Conclusion In a high proportion of patients with MSI‐high cancers and poor performance status related to end‐stage disease, salvage immunotherapy can induce potentially long‐lasting “Lazarus responses”. Immunotherapy decisions near the end‐of‐life should be carefully integrated with predictive biomarkers and with palliative care measures in the real‐world setting. Implications for Practice In this retrospective cohort study of 27 pretreated patients with microsatellite instability (MSI)‐high cancers and Eastern Cooperative Oncology Group performance status of 2 or 3 not related to comorbidities, PD‐1/PD‐L1‐based therapy induced a RECIST response in 33% of patients, with a median duration of 16.9 months, and an improvement of performance status in 52% of patients. MSI‐high status can be used in clinical practice as a tumor‐agnostic predictive biomarker to select critically ill patients with end‐stage cancers for salvage immunotherapy. This article assesses whether PD‐1/PD‐L1‐based therapy can induce Lazarus responses and clinically meaningful benefit in patients with microsatellite instability‐high end‐stage cancers and poor performance status related to high disease burden.