I’ve been on the production side of custom apparel since 2005 (I co-own a POD platform in North America — we don’t serve the EU, so no horse in your race, just production experience).
Taking your four points in order:
EU fulfillment & shipping: the single biggest thing to verify is WHERE production physically happens, not where the company is headquartered. Some providers route EU orders to partner facilities with variable quality; others run their own EU production. Ask each candidate directly: “which facility prints orders shipped to Germany, and what’s your actual measured production time there?” Then test it — order the same design twice, a Monday order and a Friday order, and time both. Stated SLAs and real ones often differ.
Textile quality: ask which specific blank brands they run (Stanley/Stella is the EU premium standard) — if a provider can’t tell you the exact garment brand and weight, that’s your answer. And always order samples washed 3+ times before judging; DTG prints in particular look identical out of the box and very different after three washes if pretreatment is inconsistent.
Returns on personalized products: under EU consumer law, personalized goods are generally exempt from the 14-day withdrawal right, but that doesn’t cover misprints or quality defects — those are on you regardless. The practical question for providers: “what’s your reprint policy and turnaround when a defect is your fault?” Good ones reprint free within days without requiring the customer to ship anything back. That policy, more than the return exemption, decides your customer-service costs.
Apps vs API: start with the app. APIs make sense once you have volume and an edge case the app can’t handle (custom packing slips, split fulfillment, etc.). One thing to check in any app: whether it does true listing sync (price/variant changes push automatically) or just order import — the difference costs you hours weekly at 100+ products.
Since you said long-term over get-rich-quick: shortlist two providers, run identical samples through both for a month, and keep the loser as your documented backup. Single-supplier dependency is the quiet killer in POD — when your only provider has a bad month, so do you.
Good luck with the evaluation!