I would go through the Redshift Spectrum best practices blog here and plan to run some tests. It is hard to quantify such metrics as every customer workload is different.
Regarding your questions:
1/ Depends on a variety of factors as noted in the best practices blog. Such as parquet file format, Snappy compression, proper partitioning on S3 to help with query access patterns/filters, type of queries such as ORDER BY, DISTINCT which cannot be pushed down to Spectrum compute layer etc. Amazon Redshift Spectrum owns managed compute layer independent of your Redshift cluster. The number of Redshift Spectrum compute nodes that a query uses depends on the Redshift node type and the overall workload. Based on the demands of your queries and Redshift cluster configuration, Redshift Spectrum scales automatically in an intelligent fashion.
2/ Same as #1
3/ Regarding query syntax difference between Athena and Redshift Spectrum, yes.
Athena's query engine is Apache Presto and hence, it follows query syntax of Apache Presto. I would refer to Presto documentation here under "SQL Language" and "SQL Statement Syntax". As far as Spectrum goes, you will find that Spectrum follows pretty much the same syntax as Redshift except things like you cannot do DML operations on Spectrum tables due to the external table.
For the second part of your question, I would make sure that customer is aware when to use Athena versus Spectrum. They are not meant to replace each other but rather meant for different workloads. Athena is more like rent-a-car for adhoc/on-demand data explorations as and when needed without needing to spin up a cluster etc. Whereas Redshift Spectrum is more like a secondary car and Redshift is your primary car. A common pattern for Redshift Spectrum is to run queries that span both the frequently accessed “hot” data stored locally in Amazon Redshift and the “warm/cold” data stored cost-effectively in Amazon S3. This pattern serves to separate compute and storage, enabling independent scaling of both to match the use case without having to pay disproportionately for value. Athena and Redshift Spectrum query optimizers are completely different. There are also differences such as you can get the same rich compliance standards of Amazon Redshift.
Is it possible to have a Redshift Spectrum schema access multiple accounts data ?asked 8 months ago
Distinguish Redshift Spectrum from Redshift CostsAccepted Answerasked 4 years ago
Can you use a Redshift Spectrum Table in a CTE?asked 10 months ago
Redshift Spectrum: no way to glance at tables with nested dataasked 4 years ago
How is data returned from Spectrum to Redshift cluster?Accepted Answerasked 3 years ago
use Redshift Spectrum to query both Redshift table and s3 fileAccepted Answerasked 2 years ago
Athena and Redshift Spectrum performance best practicesAccepted Answerasked 4 years ago
Spectrum performance caching and performanceAccepted Answerasked 4 years ago
Redshift Managed Storage vs SpectrumAccepted Answerasked 3 years ago
Redshift Spectrum with hanging select queries about 25 minutesasked 2 years ago