Digital.ai Continuous Testing could be better in certain areas, and I can share my experience-based view on what can be frustrating. One issue is device availability and queue delays during peak CI hours. Sometimes devices are busy, causing tests to queue and the pipeline to slow down unexpectedly, which is especially painful for large regression suites or tight release timelines. Improvements are needed in smarter auto-scaling of device pools and better priority-based scheduling. Additionally, execution speed variability occurs; the same test sometimes runs fast and sometimes slow, depending on device load and network latency, making results less predictable. More stable execution environments and better performance isolation per session would help. Furthermore, debugging can still be indirect; even with logs or videos, I do not fully control the device as I would with local debugging, making it hard to pause and inspect live states or reproduce edge-case issues locally. More interactive debugging and improved local reproduction tools are necessary. Cost versus usage efficiency is another area of concern, as device cloud usage can be expensive and we sometimes have idle or inefficient tests that waste money. Improvements in usage analytics and cost optimization suggestions for smart test selection to run only impacted tests are areas where I believe Digital.ai Continuous Testing could improve.
