From my experience the excel importer should be ablt to import 100k records.
Depending on your use case it might be usefull to import the file into an helper entity. That way, after the import(conversion from excel to the helper entity) you can do the processing and validation to your domain model with microflows in batches.
It depends on several variables, like how many columns are you reading and the size of your infrastructure. I've seen the performance of importing degrade with large files, so importing lines 99k-100k is much slower than liens 0-1k. This might be solved in recent Mendix versions
Several 10k lines should not be a problem, but I would just go ahead and test it. Also take the growth of data over time into account. Will you always import, say 100k lines, or will this grow over time?
I consider it a best practice to import the data in staging tables so you can add sanitize and validate the data your importing, like Maarten already mentioned.