We have a never ending poller process.Once the poller process is hitting the nod

One pattern you might want to consider is to use a Timer node as a Start Event of a process model. You can configure to start at the same interval you are using to poll currently, and after you publish the model it will automatically start new process instances at the specified time: forum.appian.com/.../Start_Event_-_Timer.html

This approach has a couple of benefits, including distribution of process instances across all of the execution engines, as well as removing the need to continually increase MAX_NODE_INSTANCES property. There are a couple of things to watch out for however:
-Use a constant to enable/disable the process timer (put in Advanced Options > Timer Conditions) to control it updating in case you ever need to disable for any reason, especially if you have system downtime.
-When there is system downtime, Appian will start instances for every timer that it missed during that interval. You should disable the timer (using the constant) during any planned downtime, and implement a workaround to handle multiple instances at the same time.

One pattern you might want to consider is to use a Timer node as a Start Event of a process model. You can configure to start at the same interval you are using to poll currently, and after you publish the model it will automatically start new process instances at the specified time: forum.appian.com/.../Start_Event_-_Timer.html

This approach has a couple of benefits, including distribution of process instances across all of the execution engines, as well as removing the need to continually increase MAX_NODE_INSTANCES property. There are a couple of things to watch out for however:
-Use a constant to enable/disable the process timer (put in Advanced Options > Timer Conditions) to control it updating in case you ever need to disable for any reason, especially if you have system downtime.
-When there is system downtime, Appian will start instances for every timer that it missed during that interval. You should disable the timer (using the constant) during any planned downtime, and implement a workaround to handle multiple instances at the same time.