I would monitor the logs in that case. Or long running queries. If most queries complete in under 5 seconds, but I start seeing a large number of transactions being open for a minute or more, I should look into that.
Monitoring for long running spikes in CPU may not be a bad idea, but that also may be normal behavior. If your system frequently runs at 20% CPU or less but has these abnormal spikes up to 100% for extended periods of time during a dump, then monitoring for that may be helpful. But it also may not. If the dump only causes a spike on a single core (for example), you may miss the spike. Or if the dump doesn't cause a spike. Generally, dumps cause memory spikes as it dumps the process and all related debugging information to disk for further analysis. But if it only needs to dump 20 MB of data, you are likely not going to catch that spike as it is pretty small.
On the other hand, if your ETL process often causes spikes after hours for hours on end, you may get a lot of false positives. False positives are very dangerous as it trains you to ignore those alerts.