It seems that in the new Zorro beta release (1.52) the issue is indeed fixed which is great. However it would be nice to know whether this bug had affected the backtest results of Z strategies.

I also think that it's worth considering whether to send an e-mail to registered users to let them know about this bug. For myself I can say that I've been running a losing strategy for 6 months because of this bug. People who don't read the forum and / or release notes that diligently might not notice and lose lot of money.

Another thing I noticed is that it makes no difference in the old version if you test with ".t1" data and TICKS flag set. And my question is: shouldn't Zorro in this particular case compute the slippage from real prices? (i.e. single ticks)

The Z strategies have not been affected. Using single ticks for determining the order fill is something different to slippage; it can be determined with the Fill mode. Slippage always picks a random price from the given range, regardless whether the simulation runs in ticks mode or not.

So if I run a simulation with BarPeriod 100 ms and set Slippage to 5, it will choose one of the following 50 bars and pick a price from the estimated price curve (which in this particular case would be probably the same as OHLC of the bar). Is that correct?

Almost. It will not pick one of the next 50 bars, but use only the next bar for determining its volatility and direction. It then picks a price that fulfils those parameters, assuming that volatility increases with the square root of time.