Application pool identity or Impersonation

Typically, you will choose different identity for worker process (or do ASP.NET impersonation) because there is need to access local/network resources that needs specific permissions. Obvious dis-advantage is that your application code may run under more permissions than it may need and thereby increasing the vulnerability against malicious attacks.

ASP.NET impersonation would have more overhead because user context needs be switched for each request. I will suggest to go with separate app pool approach - only disadvantage with app pool approach is that you have process for each one of them and so there will be overhead (from OS perspective) for each process. If your applications are smaller and don't have strong memory demands then this should not be an issue,

If you want your web services to connect to SQL via Windows authentication, you will almost certainly want to set up each application with the dedicated app pool option. This requires the least amount of setup and administration.

If you go the impersonation route, you'll need to account for the "two-hop" issue. When a user calls a web service that is using impersonation, the web service can access local resources, as that user. However, if the web service tries to connect to a non-local resource (e.g., a database running on a separate server), the result will be an authentication error. The reason is that NTLM prevents your credentials from making more than one "hop". To workaround this, you would need to use Kerberos delegation. Delegation isn't difficult to set up, but it does require Domain Admin privileges, which can make things difficult in some corporate environments.

In addition, using impersonation means that you need to manage database permissions for each user that may visit your web service. The combination of database roles and AD groups will go a long way in simplifying this, but it's an extra administrative step that you may not wish to conduct. It's also a possible security risk, as certain users may end up with privileges that are greater than your web services are anticipating.

Impersonation is useful when you need a common end user experience with other Windows services that are based on Windows security.

For example, Microsoft SharePoint servers use impersonation because you can access SharePoint document libraries with web browsers and with the standard Windows shares UI (connect / disconnect to a network share, based on the SMB protocol). To ensure security is consistent between the two, in this case, you need impersonation.

Other than this kind of scenario, impersonation is most of the time not useful (but can cost a lot in terms of scalability)