I think that raven uses TraceKit. But beyond the very low-level tools, since the entire Sentry stack is open source, I think the easiest first step is to set up Sentry on labs, use it to track errors on beta and have a look at what we're getting. Then we'll be able to look at what value each component provides to see if it's worth keeping or switching to something else.

Another good reason to go for Sentry or its components is that they're being used by other very large websites, many of which are paying customers, which means that it's getting good maintenance.

I completely missed Squash which on the whole looks less maintained and less lively than Sentry (but still looks good) but has the super-neat feature of auto-identifying culprits via git blame. I don't think it is a reason to consider them (they are standing still; Sentry moves fast, and Squash is a side project while Sentry is the core business of its developers) but maybe that feature can be extracted somehow? (Sentry has no plans for the same thing at the moment.)