Note that this parameter definitions in terms of time units refer to simplified idealized scenarios and are basically lower bounds.
The deposit related timeframes directly apply for streams where the sender account has no incoming streams and where the net flowrate is thus equal to the outgoing flowrate. If however the sender account has incoming streams, this timeframes are stretched proportionally. If for example the aggregate incoming flowrate is half of the aggregate outgoing flowrate, the time for the buffer to be consumed (critial period) doubles, as do the patrician period and the plebs period. If the aggregate incoming flowrate equals the aggregate outgoing flowrate (net flowrate = zero), those periods become potentially infinite (as long as the net flowrate doesn't change), because in that case the buffer wouldn't be consumed further, but not restored either, leaving outgoing streams critical in perpetuity.
For the TOGA exitPeriod, something similar applies - it's the lower bound for how long it would take a PIC to stream out the stake with a given exitRate, assuming nothing is added to the stake during that time. In practice, accrued liquidation rewards may be added to the stake during that time, leading to a proportional extension of the exitPeriod. In theory such added rewards could even completely offset the exitRate, leading to a net growing stake. In that case the PIC could periodically increase the exitRate (a larger stake allows for a larger exitRate) and would eventually be able to set an exitRate which leads to a shrinking stake.