Issue with infinite replay request

I believe I've seen evidence of some sort of issue with infinite replay requests.

I've attached an the annotated log of the situation. The log is from "EXCHANGE", which is a QFJ 2.1.0 application; the other party of the session, "BROKER_GATEWAY", is not QFJ. Both applications are on the same host, so the message time stamps correlate.

Essentially, what occurs in this scenario is that there is a disconnect - and while the FIX session is disconnected, EXCHANGE generates 3 execution reports (er...10, er...11, and er...12). When the session reconnects, EXCHANGE sends a logon with a sqn 12 that is greater than what BROKER_GATEWAY was expected (it was expecting sqn 11), and follows with sqn 13 (er...11) and sqn 14 (er...12)

While EXCHANGE is sending those "queued" messages, BROKER_GATEWAY issues a resend-request, for messages starting at sqn 11 and with an infinite length.

EXCHANGE responds by replaying sqns 11 (er...10), 12 (admin msg as a gap-fill), and 13 (er...11), but never replays 14 (er...12) It isn't until 30 seconds later, that another message, a heartbeat, from EXCHANGE with sqn 15 triggers a second replay request, and 14 (er...12) is replayed.

It seems as if there is some sort of race, where the resend request arrives after sqn 14 was "triggered" to be sent, but before the internal state used to prepare a resent request has been updated.