So that PNP transistor to the right of the peg seems to be OK, although in-circuit measurements were a bit inconclusive on C-E. At least not shorting, assuming that part is good.
However, the ERA-3 is officially dead. Although from the DC current draw of 32mA it looks OK (35mA according to datasheet), it measures a DC resistance of 73 ohms on it's output which can't be right, but adding up in my calculation now: 7.9V/(196+73ohms)=29.4mA + current drawn on the input. Because the datasheet of the ERA-3+ tells us the (simplified) internal circuitry. There are 2 transistors in darlington configuration inside, plus 4 biasing resistors. The 2nd transistor seems to be shorted, so we are left with 1 resistor to GND probably in the region of... 73 ohms. This makes sense now and explains the signal loss of -30dBm - -16dBm = -14dB instead of a +23dB gain. The datasheet also tells us the supply voltage to be 3.5V at 35mA, but it breaks down to 1.62V at 32mA.
Checking transistor NE021 in circuit again. Diode test: Vbe=0.78V, Vbc=0.78V, Vce=nothing, reverse polarity all 3 nothing. Looks a bit HIGH for silicium, otherwise that part seems to be OK. But why the loss in signal? At 50MHz: Base: -30dBm, Collector: -55dBm, another -25dB loss! DC voltages in operation: Vb = 0.94V, Ve=0.51V => Vbe=0.43V (unusual LOW this time!) and Vc=0.62V. I can't find any RF short on it's output, it only continues to a SMD capacitor of 10pF (as my LCR meter says). Behind that... oh, I see it now. These are 2 switching diodes, behind the upper diode some low impedance circuitry. Makes sense now, this could be about right. The 2nd transistor acts as a power amplifier it seems, so the current is not enough. Hence, my 50 MHz signal nearly disappears in the noise while a weak 49.8 MHz signal appears here now.
Will replace ERA-3 when I received it and let you know the outcome. Hope this solves the problem.