Pear (Pyrus communis L.) production in high-density orchards depends critically on the choice of a size-controlling, precocious and quality-neutral rootstock, yet most comparative information still originates from single-site trials. We evaluated four quince (Cydonia oblonga Mill.) rootstocks—‘Adams’, ‘Sydo’, QR196-9 and QR530-11—under the cultivar ‘Conference’ across five contrasting European environments of the EUFRIN thematic orchard network (Spain, Romania, United Kingdom, Poland and Norway; 41.6–59.7° N), compiling 512 plot-year observations (replicate plots of four to five trees) over up to eight seasons (2018–2025). Vegetative growth (trunk cross-sectional area, TCSA; tree height; annual increment), productivity (yield, yield efficiency, crop load, cumulative yield) and fruit quality (weight, size distribution, firmness, total soluble solids, titratable acidity and their ratio) were analyzed by two-way and per-year ANOVA, linear mixed models (LMM), variance partitioning, Finlay–Wilkinson joint regression and Wricke ecovalence stability statistics, correlation and principal component analysis. Site was the dominant source of variation for all traits (variance components up to 78% for titratable acidity, 52% for TCSA and 36% for yield), whereas the rootstock contribution, although small, was largest for TCSA (≈3%). Across environments, TCSA ranked QR196-9 > ‘Adams’ > ‘Sydo’ > QR530-11 (LMM: QR196-9 +2.77 cm2, QR530-11 −2.83 cm2 vs. ‘Adams’; p = 0.002), and QR530-11 was consistently the least productive (−1.66 kg tree−1 vs. ‘Adams’; p < 0.001). Including the trunk size at planting as a covariate did not change the QR196-9 contrast and only attenuated that of QR530-11, which had the smallest nursery trees. QR196-9 behaved as an environment-responsive genotype (regression coefficient b ≈ 1.15–1.30), whereas QR530-11 was low-yielding, with its below-average slope (b ≈ 0.82) being largely a scale effect of its smaller size (b ≈ 1.0 on the log scale); ‘Sydo’ combined intermediate performance with the highest stability. Titratable acidity was the only trait showing a significant site × rootstock interaction, confined to the Lleida site. A weighted multi-criteria index computed from the site- and year-adjusted mixed-model estimates of eight criteria, including rootstock suckering, ranked ‘Adams’ highest (7.44/10), followed by QR196-9 (5.75), ‘Sydo’ (5.64) and QR530-11 (2.76), and rootstock differences attenuated as orchards approached canopy closure. Because the four rootstocks were compared within one complete multi-site trial rather than at single sites, the mixed-model estimates are site-corrected and the resulting recommendations are transferable across environments: ‘Adams’ and ‘Sydo’ as robust general-purpose rootstocks, QR196-9 for high-potential sites and caution with QR530-11 where yield is the priority.