Collector on D604 = 4.92 volts, collector on TR616 = 0 volts and collector on TR614 = 4.23 volts before switching on.
Both the collector of TR616 and the collector of TR614 are wrong.
At initial power up, C647 will begin to charge through R658. While it is charging, the base of TR615 will remain low and the collector will stay high. A high on the collector of TR615/base of TR616 will hold the collector of TR616 low. But this condition only should exist for a few ms while C647 is charging.
When the voltage on C647 exceeds the Zener on the base of TR615, TR615 will turn on, pulling its collector low. This will do two things.... First, the collector of TR616 will go high. It will also couple a negative going pulse into the base of TR614 causing its collector to momentarily go high. The high on the collector of TR614 only lasts a short time based on the charge time of C646.
So after the power has been on for a few milliseconds, I would expect the collector of TR616 to be high (pull up at the processor) and the collector of TR614 to be low (pull down R707 near the processor).
Most questionable is the fact you are measuring the collector of TR614 as a high. There should only be a short high going pulse at that point for at power on. And by short, I mean only about 1 ms. If it stays high, I would suspect a leaky C646. But that doesn't explain the low on TR616 collector.
So I am going to ask you to recheck the voltage on the collector of TR614, TR615, and TR616 when powered on and not working, and again after it starts to work.