Partly, yes. If there are too many non-support cards in hand after Omniscience is cast then the infinite loop may not occur if Omniscience has too little shields.
The AI tends to cast cards in the order of Creatures → Supports → Spells. So if the next card in hand is Panharmonicon then it could very likely trigger the infinite loop.
Deploy the Gatewatch is 23 mana for fetching and casting 3 creatures from your deck. I’m guessing that Omniscience as a Masterpiece at 17 mana will have at least 3-4 shields.
So let’s assume Omniscience has 3 shields with a hand of only Omniscience and Panharmonicon, ie. our best case scenario.
OM is cast. OM(3 shields)
Pan is cast for free. OM(2) Pan(4) Support fetched to hand
If the fetched support is OM: OM(5) Pan(3)
If the fetched support is Pan: OM (1) Pan (7), 1 more chance to fetch OM
From this example, you have 2 chances to fetch Omniscience as the next support card from Panharmonicon firing its ability. As the only 2 supports in your deck it’s a 50-50 chance, so the odds are good. And it’ll snowball from there as Panharmonicon helps you to keep fetching extra cards whereas Omniscience casts them for free and keeps building its shields.
On the other hand with 1 additional non-support card in hand (which isn’t Whir of Invention 
), you get only 1 chance to cast a free copy of Omniscience since the free casts become Panharmonicon → Non-support card → Card fetched by the casting of Panharmonicon → Fetched non-support → Fetched support → …
That gives you an average of 1 Omniscience in 4 cards which will not be sustainable for an infinite loop if Omniscience only has 3 shields. It will need 4 shields to sustain this loop. But when the number of shields is only equal to the average number of cards fetched and cast before an Omniscience appears, it’s a fragile loop. So I guess at 3 shields it actually isn’t that bad.
But think about how the AI plays a deck with Omniscience and Panharmonicon as the only supports. After creatures it starts trying to cast Panharmonicon or Omniscience. Once Panharmonicon enters the field it fetches a support to your hand. The next card the AI chooses to charge will be a creature or a support.
If an Omniscience then gets cast we can assume that there were no more creatures in hand* and there will likely be a series of supports** that will get cast next, so your proportion of supports in hand is likely to be higher than 1-1. A ratio of 2 supports to 1 non-support will allow for Omniscience to be cast on average once every 3 cards which supports a fragile loop at 3 shields.
*or a 4th or more creature which cannot be summoned to the battlefield due to lack of space
**or at least one from Panharmonicon’s initial fetch
As a bonus, Whir of Invention helps tip the balance slightly. It adds a support summon on a non-support fetch. So you’ve a deck in Standard that could form an infinite loop.
And the deck is playable from a human perspective because you can order your cards to avoid the infinite loop. Whir of Invention and Panharmonicon to try to fetch Omniscience to cast free stuff is also perfectly legit.
Tl;dr Omniscience at 3 shields makes it hard to trigger an infinite loop with only Panharmonicon. Omniscience at 4+ shields with Panharmonicon and Whir of Invention makes an infinite loop much more likely.