[CLOSED] Wildly varying performance (T3D)

The problem I am seeing is the same game (absolutely no changes to code at all) will, on the same machine, give wildly differing performance on different runs. I've been noticing this for a while now, but haven't been able to derive any context as to what might be impacting things. In fact for a while I'd just put it down to using an old (& well in need of an upgrade) laptop. However, I'm starting to see patterns, so although the information I have is far from complete, maybe it will ring some bells for someone.

Also, I've just been able to try a second machine, and it shows the problem too - so I can eliminate it being something oddball about one machine.

First machine is an old dual core laptop, plain vanilla graphics card. Normally shows ~15 FPS (not good, but good enough to develop the game logic). When things go wrong it can drop to less than 1 FPS. Second machine is brand new with a proper 3D card. Normally shows 70 - 90 FPS. When things go wrong it can drop to less than 10 FPS.

So the impact on performance when things go wrong is severe. Here are some specific observations:

It can go wrong mid game. But usually I just suddenly discover my next run of the game is in slow mode. The rest of the applications on the machine do not seem to have their performance impacted.

The amount of time since the machine was rebooted, or the quantity of other applications loaded & running, doesn't seem to impact the problem. I've had slow mode occur on the first run after a full reboot, with virtually nothing else loaded (& conversely I've had good performance after the machine hasn't een rebooted for days, and loads of other stuff loaded & running).

My machine is not out of physical memory (i.e. it's NOT the case the machine has gone into a memory/disk swap phase).

Using process explorer I can see the game exe gets 50% CPU, regardless of whether the performnce problem is currently showing up.

Process explorer shows no new processes (or any unexpected processes at all) when in poor performance mode (i.e. it doesn't look like something has loaded itself & started doing something behind the scenes to steal my CPU).

I can abandon the game during poor performance. Come back some time later, having (so far as I am aware) done nothing to the machine, and normal good performance is back.

Due to a somewhat dodgy internet connection I have observed that the start of poor performance can be simultaneous with the restortion of an internet connection (although it can also start at other times). I've seen this twice now - althouh I accept it may just have been a coincidence. There is a full up-to-date anti virus suite (before someone asks). (I'm not using network in my game).

I have very little regularly running script. The majority is event driven, there are at most two tick driven bits of script (by using schedule to get a function to repeat itself) - one is set to run every 250ms, the other every 500ms. Both are very tight with very few actual lines of script code per execution iteration.

I have one piece of event driven script code which generates two new TSStatic models one after the other - so one would expect to see these appear on screen virtually simultaneously. Yet I've just caught the poor performance mode take over one second between the two models appearing (of course I dont know if the delay is due to the scripting, or the model creation itself, being CPU starved).

When the game is suffering from poor performance, the world editor seems OK - so it's not the quantity/complexity of the models in my world on its own which is slowing down frame rendering - something else must be implicated.

I know there's not a lot to go on - but has anyone else noticed anything similar? Better still, any suggestions as to what it might be?

If I was to hazard a guess, it's as if something is in a CPU-intensive loop pending some resource availability.