Performing large OCR batches or other long-running background tasks using Connect currently feels like "go for it and hopefully everything will work as expected" 😉 ... most of the time, it's exactly what happens, but sometimes it's not and here some improvements could help to make it less magic...
Pause and/or Stop background tasks already "in progress"
Adjust priority of background tasks, to let single OCR jobs run while having a large batch in parallel
Integrate batch results like "Feature -> OCRed" after a configurable item count, e.g. after 1.000 or 10k items
Run OCR background tasks on a(ll) Node(s) instead of the main Connect server
Stephan