We are working on a new version that will improve scalability as well as being more clear on progress status.
There are workarounds possible with the current version, to process the NCM nodes by small batches (which makes the processing smaller and more predictable), but they require a bit of configuration.
There are no real guidelines, other that one node can typically take up to 5-7 min to process depending on several factors, including but not limited to Cisco's API load.