I am trying to understand, possibly in an intuitive way, why would OQAM/FBMC when used with a Phydyas filter lead to a signal that does not need any cyclic prefix just like in any OFDM signal also created by an IFFT If the cyclic prefix is absent how would ISI be avoided?
Since FBMC doesn't depend on the channel acting as circular convolution on the transmit signal like OFDM does, it makes no sense to use a cyclic prefix. (By the way, not all OFDM is cyclic-prefix OFDM, some OFDM-schemes simply use a silent guard interval, but these are the exception.)
To battle ISI without usage of a guard interval, one has to fall back to doing an equalizer per sub-channel in FBMC (something that OFDM reduces to multiplication with the inverse of the complex channel coefficient, but bought at the expense of rate loss due to necessity of the guard interval/cyclic prefix).
However, that equalizer is typically relatively benign in complexity, since the FBMC subchannels are designed to not see very many symbols in delay spread.
That's exactly what the PHYDYAS-style FBMC does: it uses the techniques used in modern single-carrier communications per carrier, and uses OQAM to combat inter-carrier-interference, whilst employing FFT methods for the multicarrier modulation.