Bluetooth low energy (BLE) is a low power wireless technology well established in consumer products. While BLE is designed for reliable data transfer, the worst-case transmission delay is unbounded. This is problematic especially for industrial real-time applications that hinge on predictability of upper bounds on transmission delays. This paper investigates probabilistic delay bounds for different configurations of BLE. It employs a model-centric approach together with state-of-the-art probabilistic model checking techniques. We start off with measuring BLE transmission errors and latency to understand basic BLE performance characteristics. These measurements are then integrated into a probabilistic timed automata model of BLE communication which is shown to faithfully extend beyond the measured configurations. In particular, it enables us to explore the novel Isochronous channel mechanisms which recently were introduced to the Bluetooth specification with version 5.2. We show that Isochronous channels, albeit targeting the consumer market, promise significant benefits also in industrial real-time applications compared to regular BLE. These findings are made concrete by wrapping them into a typical industrial use case. We verify that BLE is reliably able to transfer safety-critical control signals within a 50 ms deadline with a probability of at least \(99.92\%\) , and better if Isochronous channels are used.