I'm very impressed to see that you've implemented a feature that I've wished for for many years: on demand attach-and-profile.
I cannot count the number of times that some process/test/application was showing bad performance, and I was thinking "if only I started this with a profiler agent".
I'm not sure if you're the first commercial vendor that provides this feature, but if you are, I commend you
The only thing I wish for is a more high-level presentation of probes.
Something like spring insight, showing top-level HTTP calls, with child tree elements for transactions, in turn containing individual database events.
Note that top-level nodes could also be executing integration tests. Child elements could be JMS operations, or outgoing HTTP/REST calls.
Such a high level view - together with the new live attach capability - would cause me to have the profile GUI running all the time, I think.
Add a view for displaying response times of an HTTP endpoint (averages, worst n%), and I'd happily give you my credit card for a license upgrade already.
Of course, such a more monitoring/insight-oriented focus would be in addition (not instead of) to the old, trusted in-detail CPU/heap analysis features.
Anyway, good job so far!