We have this recurring issue that each time this particular operation A is loaded in BOI and started, within the next minute or so, the watchdog failure is triggered and as a result holds up other existing running operations. Ironically, this particular operation A didn't get hold up while the rest did.
We found that during this short period of time when the operation A gets loaded in BOI, the free time in this controller (to which the phases of this operation A are assigned to) gets extremely low. But we just couldn't figure out how this could lead to a hold-up of the other operations/phases which are assigned to other controllers. Neither could we find anything wrong with the operation A nor the phases that could possibly 'eat up' its own controller resources.
The operation A contains only 2 phases, both phases are, I would say, moderate in size (we have bigger, complicated phases requiring the acquisition of more EMs but these phases do not pose watchdog issues), coding within the phases are pretty standard with the rest, no dynamic references are used in these 2 phases too.
What else are we missing here? Any other contributing factors that can possibly take up controller resources?
Have you tried running the phases "Always in Memory" in the phase instances? This would eliminate some of the "Cha Chunk" every time a phase loads and has to bind with the aliases, etc. You can do it per instance of the phase in the properties option. Each phase eats about 300-400K of memory, so just make sure you check that if you try this option.