When guiding with PHD2 fails during a schedule, it causes a capture failure marking the capture as "Aborted". But when the scheduler tries to recover, the PHD2 guide adaptor returns the wrong value: guideInterface->call(QDBus::AutoDetect, "getStatus") returns "no error" despite PHD has stopped guiding. As a consequence, the scheduler simply executes the next capture sequence and does not restarts guiding.
This should be fixed in 2.9.8 or GIT now if you want to test that.
Tested it on commit 9c7bebf8421ce49194ebed7ecdaa426ac143032d. Looks good!