I don't mind CAPTCHAs to slow down bots (or, more likely, have them skip that particular target). On the other hand, in my opinion assigning different priorities isn't worth the effort or at least way outside the scope of this example. To find out whether a particular request is a login, the example code needs to go through the routing table in the application, so you've already passed any front ends which might be able to schedule requests according to some priority. Running another backend layer just for logins seems like over-engineering.

In general, any proactive measures against bad bot behavior are an uphill struggle, even more so in an open source environment. Bot developers are at an advantage: They see your code and can design the attack methods accordingly. It is this imbalance why I recommend security logging by the application, even in a simple example like this. The application can help to detect the attack pattern, or to leave that job to security specialists, but only if the data are made available by the application. In particular, making the log entries machine readable is something the application must take care of.