When propagating with an event locator configured, GMAT buffers the ephemeris to an SPK file in the background. When propagating from an SPK file, this data already exists in the kernel pool, but GMAT still buffers to a duplicate ephemeris, then performs event location from that.
This isn't a performance issue, rather a disk space issue for really large SPK files. And it's just kind of ugly behavior.
This doesn't seem trivial because the ephemeris buffering code is hooked up before GMAT knows about any propagators, so it can't perform a check.
The workaround is to not propagate the spacecraft. Event location will still work, but the user may want to propagate for other reasons (e.g. graphics. parameters, etc.).