There are not any hard limits for API calls, however the load balancers will attempt to give everyone QoS.
There is some speed limitation in the api IMO.
Using single threaded restApi i got ~200-300kb/s max speed on the 50k max limit of historical data tick requests .
With many more client threads the speed rised to ~2MB/s but not linearly.
As far as constructing bars without those conditions, they would be quite far off.
Its true.
Below is comparison of polygon bar data(constructed by you) vs bid/ask data constructed from nbbo tick.
While constructing i didnot go by those rules earlyer mentioned.
Later might do chart comparison to compare if i get same result as your constructed data if constructing using the rules.
But in my comparisions i saw some strange data and since i am no expert on valid stock data i have many questions(with images to explain).
The symbol for comparison is AAPL , 10 minute bars rendered to both synchronized charts.
(left is "aggregate bars" constructed by polygon, right is constructed from "nbbo tick" of polygon(without going by guidelines))

Why some bars have just 1 tick?

This is 2 charts on top of eachother.(white below is left chart)

Are these premarket gaps in as high volume stock as AAPL(is it premarket)?

2

What causes that upward spike is there(on left chart)?

5
What causes those 0 priced bid bars in first bars of many days?

8

9

10- looks similar

Seems like the bid bar is higher than highest bid tick in tick data.

12-Strange split
AAPL was first symbol checked and didnot take long to spot strange differences.
It seems complex to me why stock prices of same provider dont match and need special conversion methods etc.