A nonretentive on-delay timer loses elapsed time when its input becomes false. Two short true periods cannot be added into one complete delay. Watch ET as well as Q when diagnosing a flickering permissive. A new request after an interruption must start its timing from zero, even if the earlier interval almost finished.
Interrupt a request one scan before completion and inspect the elapsed time. It must become zero when the request falls. Start again and stop just before the new preset boundary. A solution that preserves the earlier elapsed interval would produce a premature output, even if its initial uninterrupted delay was correct.
Read the tag list and the requested sequence before editing.